DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Election/Restrictions
Applicant’s election without traverse of 6-8 in the reply filed on12/2/2025 is acknowledged.
Double Patenting
Examiner notes that terminal disclaimer has been filed 06/05/2026 and thus overcomes the rejection made in the previous office action.
Priority
Applicant claims the benefit of prior-filed a U.S. Patent Application No. 14/907,503, filed January 25, 2016, which is the U.S. National Stage of International Application No. PCT/JP2015/068459, filed June 26, 2015, which in turn claims the benefit of Japanese Patent Application No. 2015-115532, filed June 8, 2015. Applicant claim for benefit is acknowledged.
Drawings
The drawings were received on 09/01/2022. These drawings are acceptable.
Information Disclosure Statement
The information disclosure statement (IDS) submitted on the following date(s): 9/01/2022 has (have) been considered by the examiner.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 28 and 31 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Regarding claims 28 and 31, the claims recites abbreviations that render the claims indefinite because the words/phrases associated with the claimed abbreviations are not provided in the claim language.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim 2, 8-12, 14-19, 22-24 and 26-31 are rejected under 35 U.S.C. 103 as being unpatentable over Sinyavskiy et al. (US 20130325768, hereinafter ‘Sin’) in view of Vanhoucke et al. (US 9460711, hereinafter ‘Van’).
Regarding independent claim 2, Sin teaches a system for execution of a neural network, comprising: (in [0185] FIGS. 6A-6B illustrate exemplary implementations of reconfigurable partitioned neural network apparatus comprising generalized learning framework, described above. The network 600 of FIG. 6A may comprise several partitions 610, 620, 630, comprising one or more of nodes 602 receiving inputs 612 {X} via connections 604, and providing outputs via connections 608. And in [0278] Generalized learning methodology described herein may enable different parts of the same network to implement different adaptive tasks (as described above with respect to FIGS. 5B-5C). The end user of the adaptive device may be enabled to partition network into different parts, connect these parts appropriately, and assign cost functions to each task (e.g., selecting them from predefined set of rules or implementing a custom rule)...)
at least one first device located at a first location, the at least one first device comprising first processor circuitry, first memory circuitry, and first communication interface circuitry; and at least one second device located at a second location different from the first location, the at least one second device comprising second processor circuitry, second memory circuitry, and second communication interface circuitry, wherein the at least one first device and the at least one second device are configured to: communicate with each other over a communication network distinct from the neural network; (in As depicted in Fig. 6A-B and in [0185] FIGS. 6A-6B illustrate exemplary implementations of reconfigurable partitioned neural network apparatus comprising generalized learning framework, described above. The network 600 of FIG. 6A may comprise several partitions [at least one first device located at a first location, the at least one first device comprising first processor circuitry, first memory circuitry, and first communication interface circuitry; and at least one second device located at a second location different from the first location, the at least one second device comprising second processor circuitry, second memory circuitry, and second communication interface circuitry as one of several partitions] 610, 620, 630, comprising one or more of nodes 602 receiving inputs 612 {X} via connections 604, and providing outputs via connections 608… [0188] Different partitions of the network 600 may be configured, in some implementations, to perform specialized functionality [Claimed first device and at least one second device as one of several partitions]. By way of example, the partition 610 may adapt raw sensory input of a robotic apparatus to internal format of the network (e.g., convert analog signal representation to spiking) using for example, methodology described in U.S. patent application Ser. No. 13/314,066, filed Dec. 7, 2001, entitled "NEURAL NETWORK APPARATUS AND METHODS FOR SIGNAL CONVERSION", incorporated herein by reference in its entirety. The output {Y1} of the partition 610 may be forwarded to other partitions, for example, partitions 620, 630, as illustrated by the broken line arrows 618, 618_1 in FIG. 6A [depicted partitions as claimed the at least one first device comprising first processor circuitry, first memory circuitry, and first communication interface circuitry … ; and at least one second device located at a second location different from the first location, the at least one second device comprising second processor circuitry, second memory circuitry, and second communication interface circuitry]. The partition 620 may implement visual object recognition learning that may require training input signal y.sup.d.sub.j(t) 616, such as for example an object template and/or a class designation (friend/foe). The output {Y2}) of the partition 620 may be forwarded to another partition (e.g., partition 630) as illustrated by the dashed line arrow 628 in FIG. 6A. The partition 630 may implement motor control commands required for the robotic arm to reach and grasp the identified object, or motor commands configured to move robot or camera to a new location, which may require reinforcement signal r(t) 614. The partition 630 may generate the output {Y} 638 of the network 600 implementing adaptive controller apparatus (e.g., the apparatus 520 of FIG. 5). The homogeneous configuration of the network 600 [wherein the at least one first device and the at least one second device are configured to: communicate with each other over a communication network distinct from the neural network;], illustrated in FIG. 6A, may enable a single network comprising several generalized nodes of the same type to implement different learning tasks (e.g., reinforcement and supervised) simultaneously. Examiner notes that the communication network depicted the different claimed interfaces for sending control signals between the circuit partitioned.
Examiner notes that communicating data over a network requires a communication network distinct from a machine learning model, i.e. a model and a communication network would be understood as distinct components associated with their respective functions. Also see Sin [0067] As used herein, the term "bus" is meant generally to denote all types of interconnection or communication architecture that is used to access the synaptic and neuron memory. The "bus" may be optical, wireless, infrared, and/or another type of communication medium. The exact topology of the bus could be for example standard "bus", hierarchical bus, network-on-chip, address-event-representation (AER) connection, and/or other type of communication topology used for accessing, e.g., different memories [the at least one first device comprising first processor circuitry, first memory circuitry, and first communication interface circuitry … ; and at least one second device located at a second location different from the first location, the at least one second device comprising second processor circuitry, second memory circuitry, and second communication interface circuitry] in pulse-based system [wherein the at least one first device and the at least one second device are configured to: communicate with each other over a communication network distinct from the neural network]… [0152] Different partitions of the network 600 may be configured, in some implementations, to perform specialized functionality. By way of example, the partition 610 may adapt raw sensory input of a robotic apparatus to internal format of the network (e.g., convert analog signal representation to spiking) using for example, methodology described in U.S. patent application Ser. No. 13/314,066, .. The output {Y1} of the partition 610 may be forwarded to other partitions, for example, partitions 620, 630, as illustrated by the broken line arrows 618, 618_1 in FIG. 6A [wherein the at least one first device and the at least one second device are configured to: communicate with each other over a communication network distinct from the neural network]. The partition 620 may implement visual object recognition learning that may require training input signal y.sup.d.sub.j(t) 616, such as for example an object template and/or a class designation (friend/foe). The output {Y2}) of the partition 620 may be forwarded to another partition (e.g., partition 630) as illustrated by the dashed line arrow 628 [wherein the at least one first device and the at least one second device are configured to: communicate with each other over a communication network distinct from the neural network] in FIG. 6A….[0153] In one or more implementations, the input 612 may comprise input from one or more sensor sources (e.g., optical input {Xopt} and audio input {Xaud}) with each modality data being routed to the appropriate network partition, for example, to partitions 610, 630 of FIG. 6A, respectively [the at least one first device comprising first processor circuitry, first memory circuitry, and first communication interface circuitry … ; and at least one second device located at a second location different from the first location, the at least one second device comprising second processor circuitry, second memory circuitry, and second communication interface circuitry, wherein the at least one first device and the at least one second device are configured to: communicate with each other over a communication network distinct from the neural network].
PNG
media_image1.png
700
520
media_image1.png
Greyscale
)
and execute the neural network collectively by executing a first part and a second part of the neural network, (in [0149] FIGS. 6A-6B illustrate exemplary implementations of reconfigurable partitioned neural network apparatus comprising generalized learning framework, described above. The network 600 of FIG. 6A may comprise several partitions 610, 620, 630, comprising one or more of nodes 602 receiving inputs 612 {X} via connections 604, and providing outputs via connections 608 [and execute the neural network collectively by executing a first part and a second part of the neural network].)
wherein the first memory circuitry is configured to store the first part of the neural network, and the second memory circuitry is configured to store the second part of the neural network distinct from the first part, wherein the first processor circuitry is configured to execute the first part of the neural network stored in the first memory circuitry, and the second processor circuitry of is configured to execute the second part of the neural network stored in the second memory circuitry, (as depicted in Fig. 6A and 6B and in [0174] At step 834, the controller partitions (e.g., the partitions 520_6, 520_7, 520_8, 520_9, of FIG. 5B, and/or partitions 610, 620, 630 of FIG. 6A [wherein the first memory circuitry is configured to store the first part of the neural network, and the second memory circuitry is configured to store the second part of the neural network distinct from the first part, as controller circuitry partitions wherein the first processor circuitry is configured to execute the first part of the neural network stored in the first memory circuitry, and the second processor circuitry of is configured to execute the second part of the neural network stored in the second memory circuitry,) may be configured in accordance with the learning rules (e.g., supervised, unsupervised, reinforcement, and/or any combination thereof) corresponding to the task received at step 832. Subsequently, individual partitions may be operated according to, for example, the method 800 described with respect to FIG. 8A… And in [0217] In one or more implementations, the generalized learning apparatus of the disclosure may be implemented as a software library configured to be executed by a computerized neural network apparatus (e.g., containing a digital processor). In some implementations, the generalized learning apparatus may comprise a specialized hardware module (e.g., an embedded processor or controller). In some implementations, the spiking network apparatus may be implemented in a specialized or general purpose integrated circuit (e.g., ASIC, FPGA, and/or PLD)… [0218] Advantageously, the present disclosure can be used to simplify and improve control tasks for a wide assortment of control applications including, without limitation, industrial control, adaptive signal processing, navigation, and robotics [wherein the first processor circuitry is configured to execute the first part of the neural network stored in the first memory circuitry, and the second processor circuitry of is configured to execute the second part of the neural network stored in the second memory circuitry]. Exemplary implementations of the present disclosure may be useful in a variety of devices including without limitation prosthetic devices (such as artificial limbs), industrial control, autonomous and robotic apparatus, HVAC, and other electromechanical devices [wherein the first processor circuitry is configured to execute the first part of the neural network stored in the first memory circuitry, and the second processor circuitry of is configured to execute the second part of the neural network stored in the second memory circuitry] requiring accurate stabilization, set-point control, trajectory tracking functionality or other types of control. Examples of such robotic devices may include manufacturing robots (e.g., automotive), military devices, and medical devices (e.g., for surgical robots)..)
and wherein the first processor circuitry is configured to cause the first communication interface circuitry to transmit first resultant data of the execution of the first part of the neural network to the at least one second device over the communication network, (in [0188] Different partitions of the network 600 may be configured, in some implementations, to perform specialized functionality. By way of example, the partition 610 may adapt raw sensory input of a robotic apparatus to internal format of the network (e.g., convert analog signal representation to spiking) using for example, methodology described in U.S. patent application Ser. No. 13/314,066, filed Dec. 7, 2001, entitled "NEURAL NETWORK APPARATUS AND METHODS FOR SIGNAL CONVERSION", incorporated herein by reference in its entirety. The output {Y1} of the partition 610 may be forwarded to other partitions [and wherein the first processor circuitry is configured to cause the first communication interface circuitry to transmit first resultant data of the execution of the first part of the neural network to the at least one second device over the communication network], for example, partitions 620, 630, as illustrated by the broken line arrows 618, 618_1 in FIG. 6A. The partition 620 may implement visual object recognition learning that may require training input signal y.sup.d.sub.j(t) 616, such as for example an object template and/or a class designation (friend/foe). The output {Y2}) of the partition 620 may be forwarded to another partition [and wherein the first processor circuitry is configured to cause the first communication interface circuitry to transmit first resultant data of the execution of the first part of the neural network to the at least one second device over the communication network] (e.g., partition 630) as illustrated by the dashed line arrow 628 in FIG. 6A…)
and the second processor circuitry is configured to execute the second part of the neural network based on the first resultant data received via the second communication interface circuitry from the at least one first device over the communication network. (in [0188] Different partitions of the network 600 may be configured, in some implementations, to perform specialized functionality. By way of example, the partition 610 may adapt raw sensory input of a robotic apparatus to internal format of the network (e.g., convert analog signal representation to spiking) using for example, methodology described in U.S. patent application Ser. No. 13/314,066, filed Dec. 7, 2001, entitled "NEURAL NETWORK APPARATUS AND METHODS FOR SIGNAL CONVERSION", incorporated herein by reference in its entirety. The output {Y1} [and the second processor circuitry is configured to execute the second part of the neural network based on the first resultant data received via the second communication interface circuitry from the at least one first device over the communication network] of the partition 610 may be forwarded to other partitions, for example, partitions 620, 630, as illustrated by the broken line arrows 618, 618_1 in FIG. 6A. The partition 620 may implement visual object recognition learning that may require training input signal y.sup.d.sub.j(t) 616, such as for example an object template and/or a class designation (friend/foe). The output {Y2}) [and the second processor circuitry is configured to execute the second part of the neural network based on the first resultant data received via the second communication interface circuitry from the at least one first device over the communication network] of the partition 620 may be forwarded to another partition (e.g., partition 630) as illustrated by the dashed line arrow 628 in FIG. 6A…)
Additionally, Van expressly teaches the device for executing the claimed partitions of the neural network as machine devices in 8:48-57: FIG. 5 is a conceptual illustration 500 of an example distributed framework. In one instance, computations performed for each node of a DNN may be distributed across several computing machines 502A-D so that responsibility for computation for different nodes is assigned to different computing machines [the at least one first device comprising first processor circuitry, first memory circuitry, and first communication interface circuitry … ; and at least one second device located at a second location different from the first location, the at least one second device comprising second processor circuitry, second memory circuitry, and second communication interface circuitry]. For instance, computation for node 504A may be performed by machine 502A while computation for node 504B may be performed by machine 502B. For connections between nodes that cross partition boundaries, values computed at the nodes may be transmitted between the computing machines 502A-D.
Van and Sin are analogous art because both involve developing information retrieval and processing techniques using machine learning systems and algorithms.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of the prior art developing information retrieval and processing techniques for neural network using a distributed computing framework, as disclosed by Van with the method of developing information retrieval and processing techniques using neural network models as disclosed by Sin.
One of ordinary skill in the arts would have been motivated to combine the disclosed methods disclosed by Van and Sin as noted above. Doing so allow for partitioning and managing the computational responsibility of neural networks for different nodes to different computing machines; and training larger models distributed models, (Van, 8:43-57).
Regarding claim 8, the rejection of claim 2 is incorporated and Sin further teaches the system according to claim 2, wherein the resultant data transmitted to the at least one second device is a characteristic vector result from the execution of the first part on the at least one first device. (in [0188] Different partitions of the network 600 may be configured, in some implementations, to perform specialized functionality. By way of example, the partition 610 may adapt raw sensory input of a robotic apparatus to internal format of the network (e.g., convert analog signal representation to spiking) using for example, methodology described in U.S. patent application Ser. No. 13/314,066, filed Dec. 7, 2001, entitled "NEURAL NETWORK APPARATUS AND METHODS FOR SIGNAL CONVERSION", incorporated herein by reference in its entirety. The output {Y1} [wherein the resultant data transmitted to the at least one second device is a characteristic vector result from the execution of the first part on the at least one first device] of the partition 610 may be forwarded to other partitions, for example, partitions 620, 630, as illustrated by the broken line arrows 618, 618_1 in FIG. 6A. The partition 620 may implement visual object recognition learning that may require training input signal y.sup.d.sub.j(t) 616, such as for example an object template and/or a class designation (friend/foe). The output {Y2}) [wherein the resultant data transmitted to the at least one second device is a characteristic vector result from the execution of the first part on the at least one first device] of the partition 620 may be forwarded to another partition (e.g., partition 630) as illustrated by the dashed line arrow 628 in FIG. 6A…; Examiner notes that the neural network processes vectors, in [0021] Learning rules used with spiking neuron networks may be typically expressed in terms of original spike trains instead of their secondary features (e.g., the rate or the latency from the last spike). The result is that a spiking neuron operates on spike train space, transforming a vector of spike trains (input spike trains) into single element of that space (output train)… ;
And in incorporated reference, in [0188] US Application No. 13/314,066 (i.e. US 20130151450, hereinafter ‘InCorpPon’), in [0010] The complexity of real neurons is highly abstracted when modeling artificial neurons. A schematic diagram of an artificial neuron is illustrated in FIG. 1. The model comprises a vector of inputs x=[x.sub.1, x.sub.2 . . . , x.sub.n].sup.T, a vector of weights w=[w.sub.1, . . . w.sub.n] (weights define the strength of the respective signals), and a mathematical function which determines the activation of the neuron's output y. The activation function may have various forms. In the simplest neuron models, the activation function is a linear function and the neuron output is calculated as:… [0136] In order to quantitatively evaluate the performance of learning, two distance measures are used. For analog signal outputs, the mean square error (MSE) between the target and output vectors is computed. )
Regarding claim 9, the rejection of claim 6 is incorporated and Sin further teaches the system according to claim 6, wherein the at least one second device is configured to transmit second resultant data of an execution of the second part on the at least one second device to the at least one first device. (in [0188] Different partitions of the network 600 may be configured, in some implementations, to perform specialized functionality. By way of example, the partition 610 may adapt raw sensory input of a robotic apparatus to internal format of the network (e.g., convert analog signal representation to spiking) using for example, methodology described in U.S. patent application Ser. No. 13/314,066, filed Dec. 7, 2001, entitled "NEURAL NETWORK APPARATUS AND METHODS FOR SIGNAL CONVERSION", incorporated herein by reference in its entirety. The output {Y1} of the partition 610 may be forwarded to other partitions, for example, partitions 620, 630, as illustrated by the broken line arrows 618, 618_1 in FIG. 6A. The partition 620 may implement visual object recognition learning that may require training input signal y.sup.d.sub.j(t) 616, such as for example an object template and/or a class designation (friend/foe). The output {Y2}) of the partition 620 may be forwarded to another partition (e.g., partition 630) as illustrated by the dashed line arrow 628 in FIG. 6A. The partition 630 may implement motor control commands required for the robotic arm to reach and grasp the identified object, or motor commands configured to move robot or camera to a new location, which may require reinforcement signal r(t) 614 wherein the at least one second device is configured to transmit second resultant data of an execution of the second part on the at least one second device to the at least one first device]...; And in [0190] The homogeneous nature of the network 600 may enable dynamic reconfiguration of the network during its operation. FIG. 6B illustrates one exemplary implementation of network reconfiguration in accordance with the disclosure. The network 640 may comprise partition 650, which may be configured to perform unsupervised learning task, and partition 660, which may be configured to implement supervised and reinforcement learning simultaneously [wherein the at least one second device is configured to transmit second resultant data of an execution of the second part on the at least one second device to the at least one first device]. The network configuration of FIG. 6B may be used to perform signal separation tasks by the partition 650 and signal classification tasks by the partition 660. The partition 650 may be operated according to unsupervised learning rule and may generate output {Y3} denoted by the arrow 658 in FIG. 6B. The partition 660 may be operated according to a combined reinforcement and supervised rule, may receive supervised and reinforcement input 656 [wherein the at least one second device is configured to transmit second resultant data of an execution of the second part on the at least one second device to the at least one first device], and/or may generate the output {Y4} 668.)
Additionally, as incorporated [0001] : U.S. patent application Ser. No. 13/XXX,XXX entitled "DYNAMICALLY RECONFIGURABLE STOCHASTIC SPIKING NETWORK APPARATUS AND METHODS", US Pub No. US 20130325775), hereinafter ‘InCorpSin’, teaches in sending two singles from a second device partition to a first device partition as depicted in 6C, and in [0159] The partition 690 may be configured to receive the output 688 [wherein the at least one second device is configured to transmit second resultant data of an execution of the second part on the at least one second device to the at least one first device] of the partition 680 and to further process it (e.g., perform adaptive control) via a combination of reinforcement and supervised learning. In one or more implementations, the learning rule employed by the partition 690 may comprise a hybrid learning rule. The hybrid learning rule may comprise reinforcement and supervised learning combination, as described, for example, by Eqn. 34 below. Operation of the partition 690 during learning in this implementation may be aided by teaching signal 694 r(t) [wherein the at least one second device is configured to transmit second resultant data of an execution of the second part on the at least one second device to the at least one first device]. The teaching signal 694 r(t) may comprise (1) supervisory signal y.sup.d(t), which may provide, for example, desired locations (waypoints) for an autonomous robotic apparatus; and (2) reinforcement signal r(t), which may provide, for example, how close the apparatus navigates with respect to these waypoints.
PNG
media_image2.png
682
530
media_image2.png
Greyscale
)
Regarding claim 10, the rejection of claim 9 is incorporated and Sin further teaches the system according to claim 9, wherein a third part of the neural network is stored in the at least one memory of the at least one first device, and the at least one processor of the at least one first device is configured to execute the third part of the neural network on the at least one first device based on at least the another resultant data of the execution of the second part on the at least one second device. (As depicted InCorpSin Fig. 6C and in [0159] The partition 690 [wherein a third part of the neural network is stored in the at least one memory of the at least one first device, … as node part of the first device in Fig.6C: 690] may be configured to receive the output 688 of the partition 680 and to further process it (e.g., perform adaptive control) via a combination of reinforcement and supervised learning. In one or more implementations, the learning rule employed by the partition 690 may comprise a hybrid learning rule. The hybrid learning rule may comprise reinforcement and supervised learning combination, as described, for example, by Eqn. 34 below. Operation of the partition 690 during learning in this implementation may be aided by teaching signal 694 r(t). The teaching signal 694 r(t) may comprise (1) supervisory signal y.sup.d(t), which may provide, for example, desired locations (waypoints) for an autonomous robotic apparatus; and (2) reinforcement signal r(t) [and the at least one processor of the at least one first device is configured to execute the third part of the neural network on the at least one first device based on at least the another resultant data of the execution of the second part on the at least one second device], which may provide, for example, how close the apparatus navigates with respect to these waypoints. [0160] The dynamic network learning reconfiguration illustrated in FIGS. 6A-6C may be used, for example, in an autonomous robotic apparatus performing exploration tasks (e.g., a pipeline inspection autonomous underwater vehicle (AUV), or space rover, explosive detection, and/or mine exploration). When certain functionality of the robot is not required (e.g., the arm manipulation function) the available network resources (i.e., the nodes 602) [and the at least one processor of the at least one first device is configured to execute the third part of the neural network on the at least one first device based on at least the another resultant data of the execution of the second part on the at least one second device] may be reassigned to perform different tasks. Such reuse of network resources may be traded for (i) smaller network processing apparatus, having lower cost, size and consuming less power, as compared to a fixed pre-determined configuration; and/or (ii) increased processing capability for the same network capacity. )
Regarding claim 11, the rejection of claim 9 is incorporated and Sin further teaches the system according to claim 9, wherein the second resultant data transmitted to the at least one first device is a characteristic vector result from the execution of the second part on the at least one second device; Examiner notes that the neural network processes vectors, in [0021] Learning rules used with spiking neuron networks may be typically expressed in terms of original spike trains instead of their secondary features (e.g., the rate or the latency from the last spike). The result is that a spiking neuron operates on spike train space, transforming a vector [wherein the second resultant data transmitted to the at least one first device is a characteristic vector result from the execution of the second part on the at least one second device] of spike trains (input spike trains) into single element of that space (output train)… ;
And in incorporated reference, noted in Sin [0188]: InCorpPon (US 20130151450) in [0010] The complexity of real neurons is highly abstracted when modeling artificial neurons. A schematic diagram of an artificial neuron is illustrated in FIG. 1. The model comprises a vector of inputs x=[x.sub.1, x.sub.2 . . . , x.sub.n].sup.T, a vector of weights w=[w.sub.1, . . . w.sub.n] (weights define the strength of the respective signals), and a mathematical function which determines the activation of the neuron's output y [wherein the second resultant data transmitted to the at least one first device is a characteristic vector result from the execution of the second part on the at least one second device]. The activation function may have various forms. In the simplest neuron models, the activation function is a linear function and the neuron output is calculated as:… [0136] In order to quantitatively evaluate the performance of learning, two distance measures are used. For analog signal outputs, the mean square error (MSE) between the target and output vectors [wherein the second resultant data transmitted to the at least one first device is a characteristic vector result from the execution of the second part on the at least one second device] is computed. )
Regarding claim 12, the rejection of claim 2 is incorporated and Sin further teaches the system according to claim 2, wherein the first part includes a first group of weights associated with layers of the first part, and the second part includes a second group of weights associated with layers of the second part. (in [0186] In one or more implementations, the nodes 602 of the network 600 may comprise spiking neurons [wherein the first part includes a first group of weights associated with layers of the first part, and the second part includes a second group of weights associated with layers of the second part] (e.g., the neurons 730 of FIG. 9, described below), the connections 604, 608 may be configured to carry spiking input into neurons, and spiking output from the neurons, respectively.... And in [0008] When the task changes, the learning rules (typically effected by adjusting the control parameters w={w.sub.i, w.sub.2, . . . , w.sub.n}) [wherein the first part includes a first group of weights associated with layers of the first part, and the second part includes a second group of weights associated with layers of the second part] may need to be modified to suit the new task. Hereinafter, the boldface variables and symbols with arrow superscripts denote vector quantities, unless specified otherwise. Complex control applications, such as for example, autonomous robot navigation, robotic object manipulation, and/or other applications may require simultaneous implementation of a broad range of learning tasks. Such tasks may include visual recognition of surroundings, motion control, object (face) recognition, object manipulation, and/or other tasks. In order to handle these tasks simultaneously, existing implementations may rely on a partitioning approach [wherein the first part includes a first group of weights associated with layers of the first part, and the second part includes a second group of weights associated with layers of the second part], where individual tasks are implemented using separate controllers, each implementing its own learning rule (e.g., supervised, unsupervised, reinforcement)… [0012] Even when a neural network is used as the computational engine for these learning tasks, individual tasks may be performed by a separate network partition that implements a task-specific set of learning rules (e.g., adaptive control, classification, recognition, prediction rules, and/or other rules)… [0107] One or more generalized learning methodologies described herein may enable different parts of the same network to implement different adaptive tasks. The end user of the adaptive device may be enabled to partition network into different parts, connect these parts appropriately, and assign cost functions to each task (e.g., selecting them from predefined set of rules or implementing a custom rule)…
Examiner notes that Sin teaches the neurons/nodes associated with claim weight for performing partitioned tasks of a neural network. Where the claimed portions including a plurality of layers as noted above and in [0185] FIGS. 6A-6B illustrate exemplary implementations of reconfigurable partitioned neural network [wherein the first part includes a first group of weights associated with layers of the first part, and the second part includes a second group of weights associated with layers of the second part] apparatus comprising generalized learning framework, described above. The network 600 of FIG. 6A may comprise several partitions 610, 620, 630, comprising one or more of nodes 602 receiving inputs 612 {X} via connections 604, and providing outputs via connections 608. [0186] In one or more implementations, the nodes 602 of the network 600 may comprise spiking neurons (e.g., the neurons 730 of FIG. 9, described below), the connections 604, 608 may be configured to carry spiking input into neurons, and spiking output from the neurons, respectively. The neurons 602 may be configured to generate responses (as described in, for example, U.S. patent application Ser. No. 13/152,105 filed on Jun. 2, 2011, and entitled "APPARATUS AND METHODS FOR TEMPORALLY PROXIMATE OBJECT RECOGNITION", incorporated by reference herein in its entirety) which may be propagated via feed-forward connections 608. [0187] In some implementations, the network 600 may comprise artificial neurons, such as for example, spiking neurons described by U.S. patent application Ser. No. 13/152,105 filed on Jun. 2, 2011, and entitled "APPARATUS AND METHODS FOR TEMPORALLY PROXIMATE OBJECT RECOGNITION", incorporated supra, artificial neurons with sigmoidal activation function, binary neurons (perceptron), radial basis function units, and/or fuzzy logic networks…
Additionally, the incorporated references teach the connections associated with a plurality of layers, [0188] Different partitions of the network 600 may be configured, in some implementations, to perform specialized functionality [wherein the first part includes a first group of weights associated with layers of the first part, and the second part includes a second group of weights associated with layers of the second part]…The homogeneous configuration of the network 600, illustrated in FIG. 6A, may enable a single network comprising several generalized nodes of the same type to implement different learning tasks (e.g., reinforcement and supervised) simultaneously.
[0185] FIGS. 6A-6B illustrate exemplary implementations of reconfigurable partitioned neural network apparatus comprising generalized learning framework, described above. The network 600 of FIG. 6A may comprise several partitions 610 620 [and the second part includes a second group of weights associated with layers of the second part, having processing layer 630 and segment portion of the input layer 614 connected to processing layer 630], 630 [wherein the first part includes a first group of weights associated with layers of the first part], comprising one or more of nodes 602 receiving inputs 612 {X} via connections 604 [wherein the first part includes a first group of weights associated with layers of the first part, having processing layer 620 and segment portion of the input layer 614 connected to processing layer 620], and providing outputs via connections 608… [0188] Different partitions of the network 600 may be configured, in some implementations, to perform specialized functionality. By way of example, the partition 610 may adapt raw sensory input of a robotic apparatus to internal format of the network (e.g., convert analog signal representation to spiking) using for example, methodology described in U.S. patent application Ser. No. 13/314,066, filed Dec. 7, 2001, entitled "NEURAL NETWORK APPARATUS AND METHODS FOR SIGNAL CONVERSION", incorporated herein by reference in its entirety. The output {Y1} of the partition 610 may be forwarded to other partitions [wherein the at least one first device is configured to transmit resultant data of an execution of the first part on the at least one first device to the at least one second device], for example, partitions 620, 630, as illustrated by the broken line arrows 618, 618_1 in FIG. 6A. The partition 620 may implement visual object recognition learning that may require training input signal y.sup.d.sub.j(t) 616, such as for example an object template and/or a class designation (friend/foe). The output {Y2}) of the partition 620 may be forwarded to another partition [wherein the at least one first device is configured to transmit resultant data of an execution of the first part on the at least one first device to the at least one second device] (e.g., partition 630) as illustrated by the dashed line arrow 628 in FIG. 6A…)
Regarding claim 14, the rejection of claim 2 is incorporated and Sin further teaches the system according to claim 2, wherein the at least one second device is a device communicating with a plurality of the first devices. (in [0188] Different partitions of the network 600 may be configured, in some implementations, to perform specialized functionality. By way of example, the partition 610 may adapt raw sensory input of a robotic apparatus to internal format of the network (e.g., convert analog signal representation to spiking) using for example, methodology described in U.S. patent application Ser. No. 13/314,066, filed Dec. 7, 2001, entitled "NEURAL NETWORK APPARATUS AND METHODS FOR SIGNAL CONVERSION", incorporated herein by reference in its entirety. The output {Y1} of the partition 610 may be forwarded to other partitions, for example, partitions 620, 630 [wherein the at least one second device is a device communicating with a plurality of the first devices], as illustrated by the broken line arrows 618, 618_1 in FIG. 6A. The partition 620 may implement visual object recognition learning that may require training input signal y.sup.d.sub.j(t) 616, such as for example an object template and/or a class designation (friend/foe). The output {Y2}) of the partition 620 may be forwarded to another partition (e.g., partition 630) as illustrated by the dashed line arrow 628 in FIG. 6A…)
Regarding claim 15, the rejection of claim 2 is incorporated and Sin further teaches the system according to claim 2, wherein the at least one first device is one of a personal computer, a tablet, a portable phone, a smartphone, a portable information terminal, or a touch pad. (in [0107] One or more generalized learning methodologies described herein may enable different parts of the same network to implement different adaptive tasks. The end user of the adaptive device may be enabled to partition network into different parts, connect these parts appropriately, and assign cost functions to each task (e.g., selecting them from predefined set of rules or implementing a custom rule)… [0109] Implementations of the disclosure may be, for example, deployed in a hardware and/or software implementation of a neuromorphic computer system. In some implementations, a robotic system [wherein the at least one first device is one of a .. portable information terminal …] may include a processor embodied in an application specific integrated circuit, which can be adapted or configured for use in an embedded application (e.g., a prosthetic device)…; And in [0281] Advantageously, the present disclosure can be used to simplify and improve control tasks for a wide assortment of control applications including, without limitation, industrial control, adaptive signal processing, navigation, and robotics. Exemplary implementations of the present disclosure may be useful in a variety of devices including without limitation prosthetic devices (such as artificial limbs), industrial control, autonomous and robotic apparatus, HVAC, and other electromechanical devices requiring accurate stabilization, set-point control, trajectory tracking functionality or other types of control. Examples of such robotic devices may include manufacturing robots (e.g., automotive), military devices, and medical devices (e.g., for surgical robots). Examples of autonomous navigation may include rovers (e.g., for extraterrestrial, underwater, hazardous exploration environment), unmanned air vehicles, underwater vehicles, smart appliances (e.g., ROOMBA.RTM.) [wherein the at least one first device is one of a personal computer, a tablet, a portable phone, a smartphone, a portable information terminal, or a touch pad], and/or robotic toys. The present disclosure can advantageously be used in other applications of adaptive signal processing systems (comprising for example, artificial neural networks), including: machine vision, pattern detection and pattern recognition, object classification, signal filtering, data segmentation, data compression, data mining, optimization and scheduling, complex mapping, and/or other applications.)
Regarding claim 16, the rejection of claim 2 is incorporated and Sin further teaches the system according to claim 2, wherein the at least one first device and the at least one second device are installed in different apparatuses. (in ([0188] Different partitions of the network 600 may be configured, in some implementations, to perform specialized functionality. By way of example, the partition 610 may adapt raw sensory input of a robotic apparatus to internal format of the network (e.g., convert analog signal representation to spiking) using for example, methodology described in U.S. patent application Ser. No. 13/314,066, filed Dec. 7, 2001, entitled "NEURAL NETWORK APPARATUS AND METHODS FOR SIGNAL CONVERSION", incorporated herein by reference in its entirety. The output {Y1} of the partition 610 may be forwarded to other partitions [wherein the at least one first device and the at least one second device are installed in different apparatuses], for example, partitions 620, 630, as illustrated by the broken line arrows 618, 618_1 in FIG. 6A. The partition 620 may implement visual object recognition learning that may require training input signal y.sup.d.sub.j(t) 616, such as for example an object template and/or a class designation (friend/foe). The output {Y2}) of the partition 620 may be forwarded to another partition [wherein the at least one first device and the at least one second device are installed in different apparatuses] (e.g., partition 630) as illustrated by the dashed line arrow 628 in FIG. 6A…; Patritions as different label devices having different task executed on each respective circuitry as depicted inf Figs 6A, and in [0281] Advantageously, the present disclosure can be used to simplify and improve control tasks for a wide assortment of control applications including, without limitation, industrial control, adaptive signal processing, navigation, and robotics [wherein the at least one first device and the at least one second device are installed in different apparatuses]. Exemplary implementations of the present disclosure may be useful in a variety of devices including without limitation prosthetic devices (such as artificial limbs), industrial control, autonomous and robotic apparatus, HVAC, and other electromechanical devices requiring accurate stabilization, set-point control, trajectory tracking functionality or other types of control. Examples of such robotic devices may include manufacturing robots (e.g., automotive), military devices, and medical devices (e.g., for surgical robots) [wherein the at least one first device and the at least one second device are installed in different apparatuses]. Examples of autonomous navigation may include rovers [wherein the at least one first device and the at least one second device are installed in different apparatuses] (e.g., for extraterrestrial, underwater, hazardous exploration environment), unmanned air vehicles, underwater vehicles, smart appliances (e.g., ROOMBA.RTM.), and/or robotic toys. The present disclosure can advantageously be used in other applications of adaptive signal processing systems (comprising for example, artificial neural networks), including: machine vision, pattern detection and pattern recognition, object classification, signal filtering, data segmentation, data compression, data mining, optimization and scheduling, complex mapping, and/or other applications.)
Examiner considers a device a type of apparatus and considered software-subroutines executed on distributed processing elements as depicted in Fig. 6A, as noted above and in [0280] In one or more implementations, the generalized learning apparatus of the disclosure may be implemented as a software library configured to be executed by a computerized neural network apparatus (e.g., containing a digital processor). In some implementations, the generalized learning apparatus may comprise a specialized hardware module (e.g., an embedded processor or controller). In some implementations, the spiking network apparatus may be implemented in a specialized or general purpose integrated circuit (e.g., ASIC, FPGA, and/or PLD). Myriad other implementations may exist that will be recognized by those of ordinary skill given the present disclosure.)
Regarding claim 17, the rejection of claim 2 is incorporated and Sin further teaches the system according to claim 2, wherein the execution of the neural network is a process utilizing the neural network. (in ([0188] Different partitions of the network 600 may be configured, in some implementations, to perform specialized functionality. By way of example, the partition 610 may adapt raw sensory input of a robotic apparatus to internal format of the network [wherein the execution of the neural network is a process utilizing the neural network] (e.g., convert analog signal representation to spiking) using for example, methodology described in U.S. patent application Ser. No. 13/314,066, filed Dec. 7, 2001, entitled "NEURAL NETWORK APPARATUS AND METHODS FOR SIGNAL CONVERSION", incorporated herein by reference in its entirety. The output {Y1} of the partition 610 may be forwarded to other partition, for example, partitions 620, 630, as illustrated by the broken line arrows 618, 618_1 in FIG. 6A. The partition 620 may implement visual object recognition learning that may require training input signal y.sup.d.sub.j(t) 616, such as for example an object template and/or a class designation (friend/foe). The output {Y2}) of the partition 620 may be forwarded to another partition (e.g., partition 630) as illustrated by the dashed line arrow 628 in FIG. 6A [wherein the execution of the neural network is a process utilizing the neural network]…; Patritions as different label devices having different task executed on each respective circuitry as depicted inf Figs 6A, and in [0281] Advantageously, the present disclosure can be used to simplify and improve control tasks for a wide assortment of control applications including, without limitation, industrial control, adaptive signal processing, navigation, and robotics [wherein the execution of the neural network is a process utilizing the neural network]. Exemplary implementations of the present disclosure may be useful in a variety of devices including without limitation prosthetic devices (such as artificial limbs), industrial control, autonomous and robotic apparatus, HVAC, and other electromechanical devices requiring accurate stabilization, set-point control, trajectory tracking functionality or other types of control. Examples of such robotic devices may include manufacturing robots (e.g., automotive), military devices, and medical devices (e.g., for surgical robots). Examples of autonomous navigation may include rovers (e.g., for extraterrestrial, underwater, hazardous exploration environment), unmanned air vehicles, underwater vehicles, smart appliances (e.g., ROOMBA.RTM.), and/or robotic toys. The present disclosure can advantageously be used in other applications of adaptive signal processing systems (comprising for example, artificial neural networks) [wherein the execution of the neural network is a process utilizing the neural network], including: machine vision, pattern detection and pattern recognition, object classification, signal filtering, data segmentation, data compression, data mining, optimization and scheduling, complex mapping, and/or other applications.)
Regarding independent claim 18, Sin method for execution of a neural network by at least one first device and at least one second device, the method comprising: (in [0185] FIGS. 6A-6B illustrate exemplary implementations of reconfigurable partitioned neural network apparatus comprising generalized learning framework, described above. The network 600 of FIG. 6A may comprise several partitions 610, 620, 630, comprising one or more of nodes 602 receiving inputs 612 {X} via connections 604, and providing outputs via connections 608..)
communicating, by the first device, with the at least one second device through a communication network; executing, by the at least one first device and the at least one second device, the neural network, wherein a first part of the neural network is executed on the first device and a second part of the neural network is executed on the at least one second device, the first device transmits resultant data of an execution of the first part on the first device to the at least one second device, and the at least one second device executes the second part on the at least one second device based on at least the resultant data of the execution of the first part (in As depicted in Fig. 6A-B and in [0185] FIGS. 6A-6B illustrate exemplary implementations of reconfigurable partitioned neural network apparatus comprising generalized learning framework, described above. The network 600 of FIG. 6A may comprise several partitions [communicating, by the first device, with the at least one second device through a communication network; executing, by the at least one first device and the at least one second device, the neural network, wherein a first part of the neural network is executed on the first device and a second part of the neural network is executed on the at least one second device, as one of several partitions of a neural network communicating through connections] 610, 620, 630, comprising one or more of nodes 602 receiving inputs 612 {X} via connections 604, and providing outputs via connections 608… [0188] Different partitions of the network 600 may be configured, in some implementations, to perform specialized functionality [wherein a first part of the neural network is executed on the first device and a second part of the neural network is executed on the at least one second device, the first device transmits resultant data of an execution of the first part on the first device to the at least one second device, and the at least one second device executes the second part on the at least one second device based on at least the resultant data of the execution of the first part]. By way of example, the partition 610 may adapt raw sensory input of a robotic apparatus to internal format of the network (e.g., convert analog signal representation to spiking) using for example, methodology described in U.S. patent application Ser. No. 13/314,066, filed Dec. 7, 2001, entitled "NEURAL NETWORK APPARATUS AND METHODS FOR SIGNAL CONVERSION", incorporated herein by reference in its entirety. The output {Y1} of the partition 610 [wherein a first part of the neural network is executed on the first device and a second part of the neural network is executed on the at least one second device, the first device transmits resultant data of an execution of the first part on the first device to the at least one second device, …] may be forwarded to other partitions, for example, partitions 620, 630, as illustrated by the broken line arrows 618, 618_1 in FIG. 6A. The partition 620 may implement visual object recognition learning that may require training input signal y.sup.d.sub.j(t) 616, such as for example an object template and/or a class designation (friend/foe). The output {Y2}) of the partition 620 may be forwarded to another partition (e.g., partition 630) as illustrated by the dashed line arrow 628 [the first device transmits resultant data of an execution of the first part on the first device to the at least one second device,] in FIG. 6A. The partition 630 may implement motor control commands required for the robotic arm to reach and grasp the identified object, or motor commands configured to move robot or camera to a new location, which may require reinforcement signal r(t) 614. The partition 630 […and the at least one second device executes the second part on the at least one second device based on at least the resultant data of the execution of the first part.] may generate the output {Y} 638 of the network 600 implementing adaptive controller apparatus (e.g., the apparatus 520 of FIG. 5). The homogeneous configuration of the network 600, illustrated in FIG. 6A, may enable a single network comprising several generalized nodes of the same type to implement different learning tasks (e.g., reinforcement and supervised) simultaneously.
PNG
media_image1.png
700
520
media_image1.png
Greyscale
)
Additionally, claimed the first device transmits resultant data of an execution of the first part on the first device to the at least one second device, and the at least one second device executes the second part on the at least one second device based on at least the resultant data of the execution of the first part., Examiner considers a device as software-subroutines executed on distributed processing elements as depicted in Fig. 6A [the first device transmits resultant data of an execution of the first part on the first device to the at least one second device, and the at least one second device executes the second part on the at least one second device based on at least the resultant data of the execution of the first part], as noted above and in [0280] In one or more implementations, the generalized learning apparatus of the disclosure may be implemented as a software library configured to be executed by a computerized neural network apparatus (e.g., containing a digital processor). In some implementations, the generalized learning apparatus may comprise a specialized hardware module (e.g., an embedded processor or controller). In some implementations, the spiking network apparatus may be implemented in a specialized or general purpose integrated circuit (e.g., ASIC, FPGA, and/or PLD). Myriad other implementations may exist that will be recognized by those of ordinary skill given the present disclosure.)
Regarding claim 19, the rejection of claim 18 is incorporated and Sin further teaches the method according to claim 18, wherein a plurality of layers of the neural network are split into at least the first part and the second part, the first part and the second part of the neural network being different layers of the neural network. (in [0185] FIGS. 6A-6B illustrate exemplary implementations of reconfigurable partitioned neural network [wherein a plurality of layers of the neural network are split into at least the first part and the second part, the first part and the second part of the neural network being different layers of the neural network] apparatus comprising generalized learning framework, described above. The network 600 of FIG. 6A may comprise several partitions 610, 620, 630, comprising one or more of nodes 602 receiving inputs 612 {X} via connections 604, and providing outputs via connections 608. [0186] In one or more implementations, the nodes 602 of the network 600 may comprise spiking neurons (e.g., the neurons 730 of FIG. 9, described below), the connections 604, 608 may be configured to carry spiking input into neurons, and spiking output from the neurons, respectively. The neurons 602 may be configured to generate responses (as described in, for example, U.S. patent application Ser. No. 13/152,105 filed on Jun. 2, 2011, and entitled "APPARATUS AND METHODS FOR TEMPORALLY PROXIMATE OBJECT RECOGNITION", incorporated by reference herein in its entirety) which may be propagated via feed-forward connections 608. [0187] In some implementations, the network 600 may comprise artificial neurons, such as for example, spiking neurons described by U.S. patent application Ser. No. 13/152,105 filed on Jun. 2, 2011, and entitled "APPARATUS AND METHODS FOR TEMPORALLY PROXIMATE OBJECT RECOGNITION", incorporated supra, artificial neurons with sigmoidal activation function, binary neurons (perceptron), radial basis function units, and/or fuzzy logic networks… [0188] Different partitions of the network 600 may be configured, in some implementations, to perform specialized functionality [wherein a plurality of layers of the neural network are split into at least the first part and the second part, the first part and the second part of the neural network being different layers of the neural network]…The homogeneous configuration of the network 600, illustrated in FIG. 6A, may enable a single network comprising several generalized nodes of the same type to implement different learning tasks (e.g., reinforcement and supervised) simultaneously.
Additionally, the incorporated references teach the connections associated with a plurality of layers, as incorporated [0001] : U.S. patent application Ser. No. 13/XXX,XXX entitled "DYNAMICALLY RECONFIGURABLE STOCHASTIC SPIKING NETWORK APPARATUS AND METHODS", US Pub No. US 20130325775), hereinafter ‘InCorpSin’, teaches in sending two singles from a second device partition to a first device partition as depicted in 6C, and in [0159] The partition 690 may be configured to receive the output 688 [wherein a plurality of layers of the neural network are split into at least the first part and the second part, the first part and the second part of the neural network being different layers of the neural network] of the partition 680 and to further process it (e.g., perform adaptive control) via a combination of reinforcement and supervised learning. In one or more implementations, the learning rule employed by the partition 690 may comprise a hybrid learning rule. The hybrid learning rule may comprise reinforcement and supervised learning combination, as described, for example, by Eqn. 34 below. Operation of the partition 690 during learning in this implementation may be aided by teaching signal 694 r(t) [wherein a plurality of layers of the neural network are split into at least the first part and the second part, the first part and the second part of the neural network being different layers of the neural network]. The teaching signal 694 r(t) may comprise (1) supervisory signal y.sup.d(t), which may provide, for example, desired locations (waypoints) for an autonomous robotic apparatus; and (2) reinforcement signal r(t), which may provide, for example, how close the apparatus navigates with respect to these waypoints.
PNG
media_image2.png
682
530
media_image2.png
Greyscale
Regarding claim 22, the rejection of claim 2 is incorporated and Sin further teaches the system according to claim 2, wherein the first part of the neural network and the second part of the neural network are different layers of the neural network. (the incorporated references teach the connections associated with a plurality of layers, as incorporated [0001] : U.S. patent application Ser. No. 13/XXX,XXX entitled "DYNAMICALLY RECONFIGURABLE STOCHASTIC SPIKING NETWORK APPARATUS AND METHODS", US Pub No. US 20130325775), hereinafter ‘InCorpSin’, teaches in sending two singles from a second device partition to a first device partition as depicted in 6C [wherein the first part of the neural network and the second part of the neural network are different layers of the neural network.], and in [0158] FIG. 6C illustrates an implementation of dynamically configured neuronal network 660. The network 660 may comprise partitions 670 [wherein the first part of the neural network and the second part of the neural network are different layers of the neural network], 680, 690 [wherein the first part of the neural network and the second part of the neural network are different layers of the neural network]. The partition 670 may be configured to process (e.g., to perform compression, encoding, and/or other processes) the input signal 662 via an unsupervised learning task and to generate processed output {Y5}. The partition 680 may be configured to receive the output 678 of the partition 670 and to further process it, e.g., perform object recognition via supervised learning. Operation of the partition 680 during learning may be aided by training signal 674 r(t), comprising supervisory signal y.sup.d(t), such as for example, examples of desired object to be recognized.
PNG
media_image2.png
682
530
media_image2.png
Greyscale
)
Regarding claim 23, the rejection of claim 2 is incorporated and Sin further teaches the system according to claim 2, wherein the at least one first device is configured to acquire sensor data to be processed by the neural network, the first part of the neural network includes an input layer of the neural network, and the second part of the neural network includes a layer later than the first part of the neural network. (the incorporated references teach the connections associated with a plurality of layers, as incorporated [0001] : U.S. patent application Ser. No. 13/XXX,XXX entitled "DYNAMICALLY RECONFIGURABLE STOCHASTIC SPIKING NETWORK APPARATUS AND METHODS", US Pub No. US 20130325775), hereinafter ‘InCorpSin’, teaches in sending two singles from a second device partition to a first device partition as depicted in 6C [wherein the first part of the neural network and the second part of the neural network are different layers of the neural network.], and in [0158] FIG. 6C illustrates an implementation of dynamically configured neuronal network 660. The network 660 may comprise partitions 670 [wherein the at least one first device is configured to acquire sensor data to be processed by the neural network, the first part of the neural network includes an input layer of the neural network having an input layer for capturing {X1} input], 680, 690 [and the second part of the neural network includes a layer later than the first part of the neural network]. The partition 670 may be configured to process (e.g., to perform compression, encoding, and/or other processes) the input signal 662 via an unsupervised learning task and to generate processed output {Y5}. The partition 680 may be configured to receive the output 678 of the partition 670 and to further process it, e.g., perform object recognition via supervised learning. Operation of the partition 680 during learning may be aided by training signal 674 r(t), comprising supervisory signal y.sup.d(t), such as for example, examples of desired object to be recognized.
PNG
media_image2.png
682
530
media_image2.png
Greyscale
)
Regarding claim 24, the rejection of claim 2 is incorporated and Sin further teaches the system according to claim 2, wherein a plurality of layers of the neural network are split into at least the first part and the second part, the first part and the second part of the neural network being different layers of the neural network. (in (the incorporated references teach the connections associated with a plurality of layers, as incorporated [0001] : U.S. patent application Ser. No. 13/XXX,XXX entitled "DYNAMICALLY RECONFIGURABLE STOCHASTIC SPIKING NETWORK APPARATUS AND METHODS", US Pub No. US 20130325775), hereinafter ‘InCorpSin’, teaches in sending two singles from a second device partition to a first device partition as depicted in 6C [wherein a plurality of layers of the neural network are split into at least the first part and the second part, the first part and the second part of the neural network being different layers of the neural network] and in [0158] FIG. 6C illustrates an implementation of dynamically configured neuronal network 660. The network 660 may comprise partitions 670, 680, 690 [wherein a plurality of layers of the neural network are split into at least the first part and the second part]. The partition 670 may be configured to process (e.g., to perform compression, encoding, and/or other processes) the input signal 662 via an unsupervised learning task and to generate processed output {Y5}. The partition 680 may be configured to receive the output 678 of the partition 670 and to further process it, e.g., perform object recognition via supervised learning. Operation of the partition 680 during learning may be aided by training signal 674 r(t), comprising supervisory signal y.sup.d(t), such as for example, examples of desired object to be recognized.
PNG
media_image2.png
682
530
media_image2.png
Greyscale
)
Regarding claim 26, the rejection of claim 10 is incorporated and Sin further teaches the system according to claim 6, wherein output data output from the second part of the neural network on the at least one second device by the execution of the second part of the neural network based on at least the first resultant data of the execution of the first part on the at least one first device is a control command for controlling a manufacturing device or a robot. (As depicted in Figs 6A-B, and in [0013] By way of illustration, consider a mobile robot controlled by a neural network [wherein output data output from the second part of the neural network on the at least one second device by the execution of the second part of the neural network based on at least the first resultant data of the execution of the first part on the at least one first device is a control command for controlling a manufacturing device or a robot], where the task of the robot is to move in an unknown environment and collect certain resources by the way of trial and error. This can be formulated as reinforcement learning tasks, where the network is supposed to maximize the reward signals (e.g., amount of the collected resource)… [0155] The PD block implementation denoted 474 may be configured to simultaneously implement reinforcement and supervised (RS) learning rules... By way of example, in some implementations reinforcement learning task may be to acquire resources by the mobile robot, where the reinforcement component r(t) provides information about acquired resources (reward signal) from the external environment, while at the same time a human expert shows the robot what should be desired output signal y.sup.d(t) to optimally avoid obstacles. By setting a higher coefficient to the supervised part of the performance function, the robot may be trained to try to acquire the resources if it does not contradict with human expert signal for avoiding obstacles.; And in [0188] Different partitions of the network 600 may be configured, in some implementations, to perform specialized functionality. By way of example, the partition 610 may adapt raw sensory input of a robotic apparatus [the execution of the first part on the at least one first device is a control command for controlling a manufacturing device or a robot] to internal format of the network (e.g., convert analog signal representation to spiking) using for example, methodology described in U.S. patent application Ser. No. 13/314,066, filed Dec. 7, 2001, entitled "NEURAL NETWORK APPARATUS AND METHODS FOR SIGNAL CONVERSION", incorporated herein by reference in its entirety. The output {Y1} of the partition 610 may be forwarded to other partitions, for example, partitions 620, 630, as illustrated by the broken line arrows 618, 618_1 in FIG. 6A [wherein output data output from the second part of the neural network on the at least one second device by the execution of the second part of the neural network based on at least the first resultant data of the execution of the first part on the at least one first device is a control command for controlling a manufacturing device or a robot]. The partition 620 may implement visual object recognition learning that may require training input signal y.sup.d.sub.j(t) 616, such as for example an object template and/or a class designation (friend/foe). The output {Y2}) of the partition 620 may be forwarded to another partition (e.g., partition 630) as illustrated by the dashed line arrow 628 in FIG. 6A. The partition 630 may implement motor control commands required for the robotic arm to reach and grasp the identified object, or motor commands configured to move robot [wherein output data output from the second part of the neural network on the at least one second device by the execution of the second part of the neural network based on at least the first resultant data of the execution of the first part on the at least one first device is a control command for controlling a manufacturing device or a robot] or camera to a new location, which may require reinforcement signal r(t) 614. The partition 630 may generate the output {Y} 638 of the network 600 implementing adaptive controller apparatus (e.g., the apparatus 520 of FIG. 5). The homogeneous configuration of the network 600, illustrated in FIG. 6A, may enable a single network comprising several generalized nodes of the same type to implement different learning tasks (e.g., reinforcement and supervised) simultaneously.)
Regarding claim 27, the rejection of claim 10 is incorporated and Sin further teaches the system according to claim 10, wherein output data, output from the third part of the neural network on the at least one first device by the execution of the third part of the neural network based on at least the second resultant data of the execution of the second part on the at least one second device, is a control command for controlling a manufacturing device or a robot. (As depicted in Figs 6A-B, and in [0013] By way of illustration, consider a mobile robot controlled by a neural network [wherein output data, output from the third part of the neural network on the at least one first device by the execution of the third part of the neural network based on at least the second resultant data of the execution of the second part on the at least one second device, is a control command for controlling a manufacturing device or a robot], where the task of the robot is to move in an unknown environment and collect certain resources by the way of trial and error. This can be formulated as reinforcement learning tasks, where the network is supposed to maximize the reward signals (e.g., amount of the collected resource)… [0155] The PD block implementation denoted 474 may be configured to simultaneously implement reinforcement and supervised (RS) learning rules... By way of example, in some implementations reinforcement learning task may be to acquire resources by the mobile robot, where the reinforcement component r(t) provides information about acquired resources (reward signal) from the external environment, while at the same time a human expert shows the robot what should be desired output signal y.sup.d(t) to optimally avoid obstacles. By setting a higher coefficient to the supervised part of the performance function, the robot may be trained to try to acquire the resources if it does not contradict with human expert signal for avoiding obstacles.; And in [0188] Different partitions of the network 600 may be configured, in some implementations, to perform specialized functionality. By way of example, the partition 610 may adapt raw sensory input of a robotic apparatus [the execution of the first part on the at least one first device is a control command for controlling a manufacturing device or a robot] to internal format of the network (e.g., convert analog signal representation to spiking) using for example, methodology described in U.S. patent application Ser. No. 13/314,066, filed Dec. 7, 2001, entitled "NEURAL NETWORK APPARATUS AND METHODS FOR SIGNAL CONVERSION", incorporated herein by reference in its entirety. The output {Y1} of the partition 610 may be forwarded to other partitions, for example, partitions 620, 630, as illustrated by the broken line arrows 618, 618_1 in FIG. 6A [wherein output data, output from the third part of the neural network on the at least one first device by the execution of the third part of the neural network based on at least the second resultant data of the execution of the second part on the at least one second device, is a control command for controlling a manufacturing device or a robot]. The partition 620 may implement visual object recognition learning that may require training input signal y.sup.d.sub.j(t) 616, such as for example an object template and/or a class designation (friend/foe). The output {Y2}) of the partition 620 may be forwarded to another partition (e.g., partition 630) as illustrated by the dashed line arrow 628 in FIG. 6A. The partition 630 may implement motor control commands required for the robotic arm to reach and grasp the identified object, or motor commands configured to move robot [wherein output data, output from the third part of the neural network on the at least one first device by the execution of the third part of the neural network based on at least the second resultant data of the execution of the second part on the at least one second device, is a control command for controlling a manufacturing device or a robot] or camera to a new location, which may require reinforcement signal r(t) 614. The partition 630 may generate the output {Y} 638 of the network 600 implementing adaptive controller apparatus (e.g., the apparatus 520 of FIG. 5). The homogeneous configuration of the network 600, illustrated in FIG. 6A, may enable a single network comprising several generalized nodes of the same type to implement different learning tasks (e.g., reinforcement and supervised) simultaneously.)
Regarding claim 28, the rejection of claim 2 is incorporated and Sin further teaches the system according to claim 2, wherein the first communication interface circuitry and the second communication interface circuitry implement at least one of a TCP/IP driver or a PPP driver. (in [0092] As used herein, the term "bus" is meant generally to denote all types of interconnection or communication architecture that is used to access the synaptic and neuron memory. The "bus" may be optical, wireless, infrared, and/or another type of communication medium [wherein the first communication interface circuitry and the second communication interface circuitry implement at least one of a TCP/IP driver or a PPP driver]. The exact topology of the bus could be for example standard "bus", hierarchical bus, network-on-chip, address-event-representation (AER) connection, and/or other type of communication topology used for accessing, e.g., different memories in pulse-based system.)
Regarding claim 29, the rejection of claim 15 is incorporated and Sin further teaches the system according to claim 15, wherein the at least one second device is a server device. (in [0109] Implementations of the disclosure may be, for example, deployed in a hardware and/or software implementation of a neuromorphic computer system. In some implementations, a robotic system may include a processor embodied in an application specific integrated circuit [wherein the at least one second device is a server device], which can be adapted or configured for use in an embedded application (e.g., a prosthetic device). And in 0093] As used herein, the terms "computer", "computing device", and "computerized device "may include one or more of personal computers (PCs) and/or minicomputers (e.g., desktop, laptop, and/or other PCs), mainframe computers, workstations, servers, personal digital assistants (PDAs), handheld computers, embedded computers, programmable logic devices, personal communicators, tablet computers, portable navigation aids, J2ME equipped devices, cellular telephones, smart phones, personal integrated communication and/or entertainment devices, and/or any other device capable of executing a set of instructions and processing an incoming data signal.)
Van teaches in 11:48-50: Computing device 900 can also be implemented as a personal computer, including both laptop computer and non-laptop computer configurations, or a server [wherein the at least one second device is a server device]… And in 12:25-34: It should be understood that arrangements described herein are for purposes of example only. As such, those skilled in the art will appreciate that other arrangements and other elements (e.g. machines [wherein the at least one second device is a server device], interfaces, functions, orders, and groupings of functions, etc.) can be used instead, and some elements may be omitted altogether according to the desired results. Further, many of the elements that are described are functional entities that may be implemented as discrete or distributed components or in conjunction with other components, in any suitable combination and location.
Regarding claim 30, the rejection of claim 2 is incorporated and Sin further teaches the system according to claim 2, further comprising at least one third device different from the at least one first device and the at least one second device, (in As depicted in Fig. 6A and in [0185] FIGS. 6A-6B illustrate exemplary implementations of reconfigurable partitioned neural network apparatus comprising generalized learning framework, described above. The network 600 of FIG. 6A may comprise several partitions 610, 620, 630, comprising one or more of nodes 602 receiving inputs 612 {X} via connections 604, and providing outputs via connections 608.)
wherein the at least one third device comprises third processor circuitry, third memory circuitry, and third communication interface circuitry, (in [0149] FIGS. 6A-6B [wherein the at least one third device comprises third processor circuitry, third memory circuitry, and third communication interface circuitry] illustrate exemplary implementations of reconfigurable partitioned neural network apparatus comprising generalized learning framework, described above. The network 600 of FIG. 6A may comprise several partitions 610, 620, 630, comprising one or more of nodes 602 receiving inputs 612 {X} via connections 604, and providing outputs via connections 608… [0188] Different partitions of the network 600 may be configured, in some implementations, to perform specialized functionality…[0192] As is appreciated by those skilled in the arts, the reconfiguration methodology described supra may comprise a static reconfiguration, where particular node populations are designated in advance for specific partitions (tasks) [wherein the at least one third device comprises third processor circuitry, third memory circuitry, and third communication interface circuitry]… [0209] At step 834, the controller partitions (e.g., the partitions 520_6, 520_7, 520_8, 520_9, of FIG. 5B, and/or partitions 610, 620, 630 of FIG. 6A) may be configured in accordance with the learning rules (e.g., supervised, unsupervised, reinforcement, and/or other learning rules) corresponding to the task received at step 832… [0276] Even when existing learning approaches employ neural networks as the computational engine, each learning task is typically performed by a separate network (or network partition) [wherein the at least one third device comprises third processor circuitry, third memory circuitry, and third communication interface circuitry] that operate task-specific (e.g., adaptive control, classification, recognition, prediction rules, etc.) set of learning rules (e.g., supervised, unsupervised, reinforcement).
wherein the third memory circuitry is configured to store the first part of the neural network, wherein the third processor circuitry is configured to execute the first part of the neural network stored in the third memory circuitry, and cause the third communication interface circuitry to transmit third resultant data of an execution of the first part on the at least one third device to the at least one second device in a second communication session over the communication network, , (in [0188] Different partitions of the network 600 may be configured, in some implementations, to perform specialized functionality [wherein the third memory circuitry is configured to store the first part of the neural network, wherein the third processor circuitry is configured to execute the first part of the neural network stored in the third memory circuitry]. By way of example, the partition 610 may adapt raw sensory input of a robotic apparatus to internal format of the network (e.g., convert analog signal representation to spiking) using for example, methodology described in U.S. patent application Ser. No. 13/314,066, filed Dec. 7, 2001, entitled "NEURAL NETWORK APPARATUS AND METHODS FOR SIGNAL CONVERSION", incorporated herein by reference in its entirety. The output {Y1} [and cause the third communication interface circuitry to transmit third resultant data of an execution of the first part on the at least one third device to the at least one second device in a second communication session over the communication network] of the partition 610 may be forwarded to other partitions for example, partitions 620, 630, as illustrated by the broken line arrows 618, 618_1 in FIG. 6A [and cause the third communication interface circuitry to transmit third resultant data of an execution of the first part on the at least one third device to the at least one second device in a second communication session over the communication network]. The partition 620 may implement visual object recognition learning that may require training input signal y.sup.d.sub.j(t) 616, such as for example an object template and/or a class designation (friend/foe). The output {Y2}) of the partition 620 may be forwarded to another partition (e.g., partition 630) as illustrated by the dashed line arrow 628 in FIG. 6A…)
and wherein the second processor circuitry of the at least one second device is configured to execute the second part of the neural network based on the third resultant data received via the second communication interface circuitry from the at least one third device in the second communication session, separately from the execution of the second part of the neural network based on the first resultant data received from the at least one first device in a first communication session over the communication network. ([0188] Different partitions of the network 600 may be configured, in some implementations, to perform specialized functionality. By way of example, the partition 610 may adapt raw sensory input of a robotic apparatus to internal format of the network (e.g., convert analog signal representation to spiking) using for example, methodology described in U.S. patent application Ser. No. 13/314,066, filed Dec. 7, 2001, entitled "NEURAL NETWORK APPARATUS AND METHODS FOR SIGNAL CONVERSION", incorporated herein by reference in its entirety. The output {Y1} of the partition 610 may be forwarded to other partitions, for example, partitions 620, 630, as illustrated by the broken line arrows 618, 618_1 [based on the third resultant data received via the second communication interface circuitry from the at least one third device in the second communication session … based on the first resultant data received from the at least one first device in a first communication session over the communication network as signals 618 and 618_1] in FIG. 6A. The partition 620 may implement visual object recognition learning that may require training input signal y.sup.d.sub.j(t) 616, such as for example an object template and/or a class designation (friend/foe). The output {Y2}) of the partition 620 [and wherein the second processor circuitry of the at least one second device is configured to execute the second part of the neural network based on the third resultant data received via the second communication interface circuitry from the at least one third device in the second communication session, separately from the execution of the second part of the neural network based on the first resultant data received from the at least one first device in a first communication session over the communication network] may be forwarded to another partition (e.g., partition 630) as illustrated by the dashed line arrow 628 in FIG. 6A. The partition 630 may implement motor control commands required for the robotic arm to reach and grasp the identified object, or motor commands configured to move robot or camera to a new location, which may require reinforcement signal r(t) 614. The partition 630 may generate the output {Y} 638 of the network 600 implementing adaptive controller apparatus (e.g., the apparatus 520 of FIG. 5). The homogeneous configuration of the network 600, illustrated in FIG. 6A, may enable a single network comprising several generalized nodes of the same type to implement different learning tasks (e.g., reinforcement and supervised))
Additionally, Van teaches wherein the at least one third device comprises third processor circuitry, third memory circuitry, and third communication interface circuitry as machine devices in 8:48-57: FIG. 5 is a conceptual illustration 500 of an example distributed framework. In one instance, computations performed for each node of a DNN may be distributed across several computing machines 502A-D so that responsibility for computation for different nodes is assigned to different computing machines [wherein the at least one third device comprises third processor circuitry, third memory circuitry, and third communication interface circuitry]. For instance, computation for node 504A may be performed by machine 502A while computation for node 504B may be performed by machine 502B. For connections between nodes that cross partition boundaries, values computed at the nodes may be transmitted between the computing machines 502A-D.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine the teachings of Van and Sin for the same reasons disclosed above.
Regarding claim 31, the rejection of claim 30 is incorporated and Sin further teaches the system according to claim 30, wherein the first communication session over the communication network and the second communication session over the communication network use at least one of TCP/IP or PPP. (in [0092] As used herein, the term "bus" is meant generally to denote all types of interconnection or communication architecture that is used to access the synaptic and neuron memory. The "bus" may be optical, wireless, infrared, and/or another type of communication medium [wherein the first communication session over the communication network and the second communication session over the communication network use at least one of TCP/IP or PPP]. The exact topology of the bus could be for example standard "bus", hierarchical bus, network-on-chip, address-event-representation (AER) connection, and/or other
Claim 13 is rejected under 35 U.S.C. 103 as being unpatentable over Sin in view of Van in further view of Sinyavskiy et al. (US 9489623, hereinafter ‘Sin_Polo’).
Regarding claim 13, the rejection of claim 2is incorporated and Sin further teaches the system according to 2, wherein the neural network to be executed by the at least one first device and the at least one second device is a neural network having been trained through a back propagation. (in [0155] The PD block implementation denoted 474 may be configured to simultaneously implement reinforcement and supervised (RS) learning rules… By way of example, in some implementations reinforcement learning task may be to acquire resources by the mobile robot, where the reinforcement component r(t) provides information about acquired resources (reward signal) from the external environment, while at the same time a human expert shows the robot what should be desired output signal y.sup.d(t) to optimally avoid obstacles. By setting a higher coefficient to the supervised part of the performance function, the robot may be trained [wherein the neural network to be executed by the at least one first device and the at least one second device is a neural network having been trained] to try to acquire the resources if it does not contradict with human expert signal for avoiding obstacles.
And where in the trained robot includes claimed neural network as a spiking neural network, in [0138] In some implementations the PD block may transmit the external signal r to the learning block (as illustrated by the arrow 404_1) so that: F(t)=r(t), (Eqn. 33) where signal r provides reward and/or punishment signals from the external environment. By way of illustration, a mobile robot, controlled by spiking neural network, may be configured to collect resources (e.g., clean up trash) while avoiding obstacles (e.g., furniture, walls). In this example, the signal r may comprise a positive indication (e.g., representing a reward) at the moment when the robot acquires the resource (e.g., picks up a piece of rubbish) and a negative indication (e.g., representing a punishment) when the robot collides with an obstacle (e.g., wall). Upon receiving the reinforcement signal r, the spiking neural network [wherein the neural network to be executed by the at least one first device and the at least one second device is a neural network having been trained] of the robot controller may change its parameters (e.g., neuron connection weights) in order to maximize the function F (e.g., maximize the reward and minimize the punishment).)
Sin does not express teach the trained spiking neural network as a neural network having been trained through a back propagation.
Sin_Polo does express teach the trained spiking neural network as a neural network having been trained through a back propagation. (in 10:25-33: FIG. 2 is a block diagram depicting a spiking neuron network configured for error back propagation [a neural network having been trained through a back propagation], in accordance with one or more implementations. The network 200 may comprise two layers of spiking neurons: layer one (or layer x) comprising neurons 202, 204; and layer two (or layer y), comprising neuron 222. The first layer neurons 202, 204 may receive input 208, 209 and communicate their output to the second layer neuron 222 via connections 212, 214. And in 2:13-22: Spiking neural networks may be utilized in a variety of applications such as, for example, image processing, object recognition, classification, robotics, and/or other. Such networks may comprise multiple nodes (e.g., units, neurons) interconnected with one another via, e.g., synapses (doublets, connections). As used herein “back propagation” is used without limitation as an abbreviation for “backward propagation of errors” which is a method commonly used for training artificial neural networks…)
Sin_Polo and Sin are analogous art because both involve developing information retrieval and processing techniques using machine learning systems and algorithms.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of the prior art for retrieving information from images by considering geometric features from the sensors as input, such volume and shape parameters of objects captured within the image data as disclosed by Sin_Polo with the method of developing information retrieval and processing techniques using a neural network model as disclosed by Sin.
One of ordinary skill in the arts would have been motivated to combine the disclosed methods disclosed by Sin_Polo and Sin as noted above. Doing so allowing for implementing backwards error propagation in distributed networks be utilized in machine learning tasks, (Sin_Polo, Abstract & 2:49-53).
Alternatively, claims 8 and 11-12 are rejected under 35 U.S.C. 103 as being unpatentable over Siny in view of Van in further view of Kasabov et al. (US 20030149676, hereinafter ‘Kasa’).
Regarding claim 8, the rejection of the claim as noted above is incorporated rejections above.
Alternatively, Kasa does expressly disclose the use of a vector data type, in [0037] FIG. 2 illustrates the computer-implemented aspects of the invention stored in memory 6 and/or mass storage 14 and arranged to operate with processor 4. The preferred system is arranged as an evolving connectionist system 20. The system 20 is provided with one or more neural network modules or NNM 22 [… a characteristic vector result from the execution of the first part on the at least one first device]. The arrangement and operation of the neural network module(s) 22 forms the basis of the invention and will be further described below… [0043] The neural network module 22 further comprises rule base layer 48 having one or more rule nodes 50. Each rule node 50 is defined by two vectors of connection weights W1(r) and W2(r) [… a characteristic vector result from the execution of the first part on the at least one first device]. Connection weight W1(r) is preferably adjusted through unsupervised learning based on similarity measure within a local area of the problem space. W2(r), on the other hand, is preferably adjusted through supervised learning based on output error, or on reinforcement learning based on output hints. Connection weights W1(r) and W2(r) are further described below.
Kasa, Van and Sin are analogous art because both involve developing information retrieval and processing techniques using machine learning systems and algorithms.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of the prior art developing information retrieval and processing techniques using neural network module forming part of an adaptive learning system based on neural network models as disclosed by Kasa with the method of developing information retrieval and processing techniques using neural network models as collectively disclosed by Van and Sin.
One of ordinary skill in the arts would have been motivated to combine the disclosed methods disclosed by Kasa, Van and Sin as noted above. Doing so allow for developing and implementing adaptive learning systems able to learn quickly from a large amount of data, adapt incrementally in an on-line mode, have an open structure so as to allow dynamic creation of new modules, memories information that can be used later, (Kasa, Abstract & 0003).
Regarding claim 11, the rejection of the claim as noted above is incorporated rejections above.
Alternatively, Kasa does expressly disclose the use of a vector data type, in [0037] FIG. 2 illustrates the computer-implemented aspects of the invention stored in memory 6 and/or mass storage 14 and arranged to operate with processor 4. The preferred system is arranged as an evolving connectionist system 20. The system 20 is provided with one or more neural network modules or NNM 22 [… a characteristic vector result from the execution of the second part on the at least one second device]. The arrangement and operation of the neural network module(s) 22 forms the basis of the invention and will be further described below… [0043] The neural network module 22 further comprises rule base layer 48 having one or more rule nodes 50. Each rule node 50 is defined by two vectors of connection weights W1(r) and W2(r) [… a characteristic vector result from the execution of the second part on the at least one second device.]. Connection weight W1(r) is preferably adjusted through unsupervised learning based on similarity measure within a local area of the problem space. W2(r), on the other hand, is preferably adjusted through supervised learning based on output error, or on reinforcement learning based on output hints. Connection weights W1(r) and W2(r) are further described below.
Kasa Van, and Sin are analogous art because both involve developing information retrieval and processing techniques using machine learning systems and algorithms.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of the prior art developing information retrieval and processing techniques using neural network module forming part of an adaptive learning system based on neural network models as disclosed by Kasa with the method of developing information retrieval and processing techniques using neural network models as collectively disclosed by Van and Sin.
One of ordinary skill in the arts would have been motivated to combine the disclosed methods disclosed by Kasa, Van and Sin as noted above. Doing so allow for developing and implementing adaptive learning systems able to learn quickly from a large amount of data, adapt incrementally in an on-line mode, have an open structure so as to allow dynamic creation of new modules, memories information that can be used later, (Kasa, Abstract & 0003).
Regarding claim 12, the rejection of the claim as noted above is incorporated rejections above.
Alternatively, Kasa does expressly disclose the use of weights, in [0037] FIG. 2 illustrates the computer-implemented aspects of the invention stored in memory 6 and/or mass storage 14 and arranged to operate with processor 4. The preferred system is arranged as an evolving connectionist system 20. The system 20 is provided with one or more neural network modules or NNM 22 […group of weights associated with layers]. The arrangement and operation of the neural network module(s) 22 forms the basis of the invention and will be further described below… [0043] The neural network module 22 further comprises rule base layer 48 having one or more rule nodes 50 [… group of weights associated with layers]. Each rule node 50 is defined by two vectors of connection weights W1(r) and W2(r) [… group of weights associated with layers]. Connection weight W1(r) is preferably adjusted through unsupervised learning based on similarity measure within a local area of the problem space. W2(r), on the other hand, is preferably adjusted through supervised learning based on output error, or on reinforcement learning based on output hints. Connection weights W1(r) and W2(r) are further described below.
Kasa, Van and Sin are analogous art because both involve developing information retrieval and processing techniques using machine learning systems and algorithms.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of the prior art developing information retrieval and processing techniques using neural network module forming part of an adaptive learning system based on neural network models as disclosed by Kasa with the method of developing information retrieval and processing techniques using neural network models as collectively disclosed by Van and Sin.
One of ordinary skill in the arts would have been motivated to combine the methods disclosed by Kasa, Van and Sin as noted above. Doing so allow for developing and implementing adaptive learning systems able to learn quickly from a large amount of data, adapt incrementally in an on-line mode, have an open structure so as to allow dynamic creation of new modules, memories information that can be used later, (Kasa, Abstract & 0003).
Response to Arguments
Applicant's arguments filed 06/05/2026 have been fully considered.
Examiner notes that terminal disclaimer has been filed 06/05/2026 and thus overcomes the rejection made in the previous office action.
Examiner notes that the interpretation under 35 USC 112(f) has been removed from the record as they no longer apply to the amended claim limitations which recite a hardware component. This also address the related rejections of claims under 35 USC 112(a) and 112(b). The rejection made in the previous office action, under 35 USC 112(a) and 112(b), have been withdrawn.
Regarding applicant remarks directed to the rejection of claims under 35 USC 102 and 103, the remarks are directed at amended limitations that have not been previously considered by the examiner. See the rejection above that address the amended claim language.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Breed (US 9102220) teaches sequential neural network models of a sequential neural network of robotic control for processing an identification neural network to determine the identification of the occupying item, the data used by the position/size determination neural network to determine the position of the occupying item, the data used by the orientation determination neural network, the data used by the position determination neural networks which may all be different from one another.
Izhikevich et al. (US 9764468): teaches the arrangements of neural network model based operations as depicted in Fig. 5:
PNG
media_image3.png
572
358
media_image3.png
Greyscale
Cosic (US 9443192): teaches distribution of neural network components as depicted in Fig. 38 and in 67:4-61: a device or system for autonomous application operating. The device or system may include a processor coupled to a memory unit. The device or system may further include an application, running on the processor, for performing operations on a computing device. The device or system may further include an interface for receiving a first instruction set and a second instruction set, the interface further configured to receive a new instruction set, wherein the first, the second, and the new instruction sets are executed by the processor and are part of the application for performing operations on the computing device. The device or system may further include a knowledgebase, neural network, or other repository configured to store at least one portion of the first instruction set and at least one portion of the second instruction set, the knowledgebase, neural network, or other repository comprising a plurality of portions of instruction sets. The device or system may further include a decision-making unit configured to compare at least one portion of the new instruction set with at least one portion of the first instruction set from the knowledgebase, neural network, or other repository. The decision-making unit may also be configured to determine that there is a substantial similarity between the new instruction set and the first instruction set from the knowledgebase, neural network, or other repository. The processor may then be caused to execute the second instruction set from the knowledgebase, neural network, or other repository. Any of the operations of the described elements can be performed repeatedly and/or in different orders in alternate embodiments. Specifically, in this example, Processor 11 can be implemented as a device or processing circuit that receives Software Application's 120 instructions, data, and/or other information from Memory 12… Acquisition and Modification Interface 110 may provide Software Application's 120 instructions, data, and/or other information to Artificial Intelligence Unit 130. Artificial Intelligence Unit 130 may learn the operation of Software Application 120 by storing the knowledge of its operation into Knowledgebase 530, Neural Network 850, or other repository. Decision-making Unit 540 may then anticipate or determine Software Application's 120 instructions, data, and/or other information most likely to be used, implemented, or executed in the future.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to OLUWATOSIN ALABI whose telephone number is (571)272-0516. The examiner can normally be reached Monday-Friday, 8:00am-5:00pm EST..
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Michael Huntley can be reached at (303) 297-4307. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/OLUWATOSIN ALABI/ Primary Examiner, Art Unit 2129
Prosecution Insights