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 .
This action is responsive to the application filed 03/21/2024. Claims 1-25 are canceled and 26-45 are presented for examination.
Priority
Applicant’s claim for the benefit of a prior filed application US 16/004671, filed 06/11/2018, is acknowledged.
Information Disclosure Statement
The information disclosure statements (IDS) submitted 06/24/2024, 10/03/2024, 02/25/2025, 06/09/2025, 06/27/2025, 10/01/2025, 12/05/2025, 06/09/2026 has 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 35-36 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.
The following claims have lack antecedent basis:
Claim 35, line 4, the phrase “the collection of windows”.
The following claims are indefinite:
Regarding claim 36, line 4, the phrase “the activity patterns” rendered the claim indefinite. It is unclear if “the activity patterns” is meant to replace “patterns of signal transmission activity”. For the purposes of examination, this limitation is interpreted as: patterns of signal transmission activity.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
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.
Claims 26-32, 35, & 37-43 are rejected under 35 U.S.C. 103 as being unpatentable over Maass et al. (“Real-Time Computing Without Stable States: A New Framework for Neural Computation Based on Perturbations”, Neural Computation 14, 2531–2560, Massachusetts Institute of Technology) (2002), hereafter Maass, in view of Giusti et al. (“Clique topology reveals intrinsic geometric structure in neural correlations”, Department of Mathematics & Center for Neural Engineering, The Pennsylvania State University, arXiv) (2015), hereafter Giusti.
Giusti was cited in the IDS submitted 06/09/2026.
Regarding independent claim 26, Maass teaches a system, comprising:
a recurrent neural network coupled to input signals, wherein the recurrent neural network is trained to process the input signals and produce a responsive output using one or more reader nodes ([Abstract] discusses readout neurons within a recurrent neural network are coupled to inputs and can learn to extract output from those inputs);
wherein the one or more reader nodes are configured to identify decisions within signal transmission activity in the recurrent neural network that is responsive to an input signal ([Sec. 1] discusses the readout neurons can identify and create definitions and decisions within the activity of input signal transmissions);
wherein each of the reader nodes is coupled to a respective of the particular collections of nodes (Maass [Sec. 5] discusses readout modules trained to fire upon detecting the occurrence of a pattern; thus, they are coupled to certain patterns and activity).
Maass does not explicitly teach the decisions reflected in occurrences of particular patterns of signal transmission activity at particular collections of nodes in the recurrent neural network.
However, Giusti teaches a system for detecting structure in neural activity by analyzing patterns of collections of nodes within a clique topology ([Pg. 3] discusses the connected subgraphs, which represent collections of nodes, can be used to detect random or geometric structure); [Pg. 15] discusses computing Betti curves by finding maximal cliques of the connected subgraphs; thus, output is reflected in the patterns of signal activity at collections of nodes of a recurrent neural network).
Because Maass teaches a recurrent neural network coupled to input signals, producing output using reader nodes, identifying decisions within signal transmission activity, and reader nodes coupled to collections of nodes; and Giusti teaches output reflected in occurrences of patterns at particular collections of nodes, accordingly, it would have been obvious to one of ordinary skill in the art to incorporate output reflected in occurrences of patterns at particular collections of nodes as taught by Giusti into Maass’ system, with a reasonable expectation of success, to teach an encoding device implemented in hardware or in a combination of hardware and software, the encoding device comprising: a recurrent neural network coupled to input signals, wherein the recurrent neural network is trained to process the input signals and produce a responsive output using one or more reader nodes, wherein the one or more reader nodes are configured to identify decisions within signal transmission activity in the recurrent neural network that is responsive to an input signal, the decisions reflected in occurrences of particular patterns of signal transmission activity at particular collections of nodes in the recurrent neural network, wherein each of the reader nodes is coupled to a respective of the particular collections of nodes. This combination would have been motivated by the desire to use clique topology to detect organization, structure, and patterns within the neural network (Giusti [Sec. 1]).
Regarding dependent claim 27, the combination of Maass and Giusti teaches the claimed invention as claimed in claim 26, including wherein each reader node activates in response to an occurrence of a pattern of activity involving the respective collection of nodes (Maass [Sec. 5] discusses readout modules trained to fire upon detecting the occurrence of a pattern, and becomes active).
Regarding dependent claim 28, the combination of Maass and Giusti teaches the claimed invention as claimed in claim 27. Additionally, Giusti teaches wherein individual reader nodes are connected to different nodes and a different number of nodes (Giusti [Pg. 15] discusses connecting and finding all maximal cliques of up to five vertices and thus, reader nodes are connected to different nodes, as well as a different number of nodes). Because Maass teaches a recurrent neural network coupled to input signals, producing output using reader nodes, identifying decisions within signal transmission activity, and reader nodes coupled to collections of nodes; and Giusti teaches individual reader nodes connected to different nodes and a different number of nodes, accordingly, it would have been obvious to one of ordinary skill in the art to incorporate individual reader nodes connected to different nodes and a different number of nodes as taught by Giusti into Maass’ system, with a reasonable expectation of success, to teach wherein individual reader nodes are connected to different nodes and a different number of nodes. This combination would have been motivated by the desire to compute Betti curves in order to summarize the topological features (Giusti [Pg. 15]).
Regarding dependent claim 29, the combination of Maass and Giusti teaches the claimed invention as claimed in claim 27. Additionally, Giusti teaches wherein individual reader nodes are connected to different links between nodes and different numbers of links between nodes (Giusti [Methods] discusses indexing graphs by edge density, wherein the number of links is varied; thus, the reader nodes connect to different links between nodes and different numbers of links between nodes). Because Maass teaches a recurrent neural network coupled to input signals, producing output using reader nodes, identifying decisions within signal transmission activity, and reader nodes coupled to collections of nodes; and Giusti teaches individual reader nodes connected to different links between nodes and different numbers of links between nodes, accordingly, it would have been obvious to one of ordinary skill in the art to incorporate individual reader nodes connected to different links between nodes and different numbers of links between nodes as taught by Giusti into Maass’ system, with a reasonable expectation of success, to teach wherein individual reader nodes are connected to different links between nodes and different numbers of links between nodes. This combination would have been motivated by the desire to organize the Betti numbers across all graphs in the order complex into Betti curves (Giusti [Pg. 10]).
Regarding dependent claim 30, the combination of Maass and Giusti teaches the claimed invention as claimed in claim 27, including wherein individual reader nodes are configured with tailored responses to identify different patterns of signal transmission activity, the tailored responses comprising different decay times in an integrate-and-fire model (Maass [Appendix B] discusses using different decay times within the integrate-and-fire model, based on the population of neurons, which is a tailored response to each pattern of signal transmission activity).
Regarding dependent claim 31, the combination of Maass and Giusti teaches the claimed invention as claimed in claim 26, including wherein the encoding device is configured to transmit or store the output of the recurrent neural network (Maass [Sec. 1] discusses the readout neurons extract and transmit information to other microcircuits; thus, the output of the recurrent neural network is transmitted).
Regarding dependent claim 32, the combination of Maass and Giusti teaches the claimed invention as claimed in claim 26. Additionally, the combination of Maass and Giusti teaches wherein content of the output of the recurrent neural network comprises activity in the recurrent neural network that matches patterns indicative of complexity in the activity, and wherein the encoding device is configured to only transmit or store activity that matches relatively more complex or higher dimensional activity (Giusti [Pg. 3] discusses the cliques and subgraphs of the network can be used to detect random or geometric structure and thus, the output of these cliques of the recurrent neural network comprises activity indicative of complexity or structure; Maass [Sec. 1] discusses the readout neurons can learn to extract and transmit information from high-dimensional transient states and transform circuit states into stable readouts; thus, only more complex or higher dimensional activity is transmitted or stored). Because Maass teaches only more complex or higher dimensional activity is transmitted or stored; and Giusti teaches output of cliques of the recurrent neural network comprises activity indicative of complexity or structure, accordingly, it would have been obvious to one of ordinary skill in the art to incorporate output of cliques of the recurrent neural network comprises activity indicative of complexity or structure as taught by Giusti into Maass’ system, with a reasonable expectation of success, to teach wherein content of the output of the recurrent neural network comprises activity in the recurrent neural network that matches patterns indicative of complexity in the activity, and wherein the encoding device is configured to only transmit or store activity that matches relatively more complex or higher dimensional activity. This combination would have been motivated by the desire to use clique topology to output and measure complexity and structure for further transmitting, analysis, or storing (Giusti [Pg. 3]).
Regarding dependent claim 35, the combination of Maass and Giusti teaches the claimed invention as claimed in claim 26, including wherein the input signal comprises a continuous stream of information that is injected into one or more nodes or one or more links of the neural network over a period of time, and wherein dividing the activity in the recurrent neural network into the collection of windows comprises subdividing a duration of the injection into windows during which the activity displays variable complexities (Maass [Abstract] discusses the goal of the system is to inject a continuous stream of multimodal input into one or nodes and process it using integrate-and-fire neurons in real time; Maass [Sec. 10] discusses subdividing a continuous injection, with 2 seconds each, into a time series, where readouts extract time-varying information and complexities; thus, the stream is analyzed in windows across its duration).
Regarding claims 37-41, they are method claims that are substantially the same as the machine of claims 26-30, respectively. Therefore, claims 37-41 are rejected for the same reasons as claims 26-30, respectively.
Regarding dependent claim 42, the combination of Maass and Giusti teaches the claimed invention as claimed in claim 37. Additionally, Giusti teaches wherein the training dataset comprises a plurality of representations of topological structures in patterns of signal transmission activity (Giusti [Pg. 15] discusses collecting all maximal cliques up to five vertices and using it as a training dataset; thus, the cliques represent topological structures in patterns). Because Maass teaches a recurrent neural network coupled to input signals, producing output using reader nodes, identifying decisions within signal transmission activity, and reader nodes coupled to collections of nodes; and Giusti teaches the training dataset comprises a plurality of representations of topological structures in patterns of signal transmission activity, accordingly, it would have been obvious to one of ordinary skill in the art to incorporate the training dataset comprising a plurality of representations of topological structures in patterns of signal transmission activity as taught by Giusti into Maass’ system, with a reasonable expectation of success, to teach wherein the training dataset comprises a plurality of representations of topological structures in patterns of signal transmission activity. This combination would have been motivated by the desire to compute the clique topology based on a training dataset that contains the features of the topological structures (Giusti [Pg. 14-16]).
Regarding dependent claim 43, the combination of Maass and Giusti teaches the claimed invention as claimed in claim 42, including wherein the training set further comprises a plurality of input vectors each corresponding to a respective of the plurality of representations; and training the recurrent neural network comprises training the recurrent neural network using each of the plurality of representations as a target answer vector (Maass [Appendix B] discusses using readout elements, which equate to the plurality of representations, as perceptrons to train the network during a simulation, by acting as a target answer vector).
Claims 33-34, & 36 are rejected under 35 U.S.C. 103 as being unpatentable over Maass, in view of Giusti, as applied in claim 26, and further in view of Masulli et al. (“The Topology Of The Directed Clique Complex As A Network Invariant”, Neuroheuristic Research Group, Faculty of Business and Economics (HEC), arXiv) (2016), hereafter Masulli.
Masulli was cited in the IDS submitted 06/09/2026.
Regarding dependent claim 33, the combination of Maass and Giusti teaches the claimed invention as claimed in claim 26, including:
determine, for each window in the collection of windows, a complexity of the patterns in the activity in the network, wherein the complexity represents a likelihood that an ordered pattern of activity arises within the window (Giusti [Pg. 3 & 10] discusses the arrangement of cliques can be used to detect structure, which constitutes the complexity of each pattern; computing Betti curves across the graphs by edge density is a topological measure of how likely the clique structure will reflect that complexity, and it is computed at each density threshold);
and time a reading of an output from the recurrent neural network based on the timings of the patterns in the activity in the network that have the distinguishable complexity (Maass [Sec. 6] discusses reading an output of a recurrent neural network based on a certain time such as t = 1000 ms).
The combination of Maass and Giusti does not explicitly teach wherein the encoding device is configured to: divide activity in the recurrent neural network that is responsive to a respective input signal into a collection of windows; identify, in each window in the collection of windows, patterns in the activity in the recurrent neural network; determine timings of patterns in the activity in the network that have a distinguishable complexity.
However, Masulli teaches organizing neurons into layers to study the topological features of a clique complex and determining distinguishable complexity ([Pg. 9] discusses organizing neurons into layers and steps, which constitutes windowing activity in response to input signals; [Abstract] discusses when the network evolution occurs through each window, the system’s goal is to show the topological features of the clique complex; [Abstract & Pg. 9-10] discusses monitoring patterns using activity-dependent plasticity at different time steps and strengthening or pruning connections; thus, the network inherently records distinguishable changes within the topology).
Because the combination of Maass and Giusti teaches determining a complexity of the patterns in the activity in the network and timing a reading of an output based on the timings of the patterns in the activity of the network that have distinguishable complexity; and Masulli teaches dividing activity into a collection of windows, identifying patterns in the activity within each window, and determining timings of patterns that have a distinguishable complexity, accordingly, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate dividing activity into a collection of windows, identifying patterns in the activity within each window, and determining timings of patterns that have a distinguishable complexity as taught by Masulli, into the combination of Maass and Giusti’s system, with a reasonable expectation of success, to teach wherein the encoding device is configured to: divide activity in the recurrent neural network that is responsive to a respective input signal into a collection of windows; identify, in each window in the collection of windows, patterns in the activity in the recurrent neural network; determine, for each window in the collection of windows, a complexity of the patterns in the activity in the network, wherein the complexity represents a likelihood that an ordered pattern of activity arises within the window; determine timings of patterns in the activity in the network that have a distinguishable complexity; and time a reading of an output from the recurrent neural network based on the timings of the patterns in the activity in the network that have the distinguishable complexity. This combination would have been motivated by the desire to define an invariant of directed networks, the network degree invariant, which is constructed by computing the topological invariant on a sequence of sub-networks filtered by the minimum in- or out-degree of the nodes (Masulli [Abstract]).
Regarding dependent claim 34, the combination of Maass, Giusti, and Masulli teaches the claimed invention as claimed in claim 33, including wherein the input signal comprises a discrete injection event, wherein information is injected into one or more nodes or one or more links of the recurrent neural network, and wherein dividing the activity in the recurrent neural network into the collection of windows comprises subdividing a time between injection and a return to a quiescent state of the neural network into a number of periods during which the activity displays variable complexities (Maass [Sec. 2] discusses injection events such as sounds or noise, provides input where information is injected into one or more nodes of the recurrent neural network; Masulli [Pg. 9] discusses organizing neurons into layers and steps, which constitutes windowing activity in response to input signals; Maass [Sec. 5] discusses tracking liquid-state evolution as a function of time following an input, with complexity changing as memory decays; thus, the network subdivides a time between injection and a return to a quiescent state of the neural network into a number of periods during which the activity displays variable complexities).
Regarding dependent claim 36, the combination of Maass and Giusti teaches the claimed invention as claimed in claim 26, including wherein the activity patterns that are identified are cliques in the functional graph of the neural network (Giusti [Pg. 3] discusses the cliques in the functional graph can be used to detect features, structure, and activity patterns).
The combination of Maass and Giusti does not explicitly teach wherein identifying decisions in signal transmission activity in the recurrent neural network that is responsive to the input signal comprises treating a functional graph of the recurrent neural network as a topological space with nodes as points.
However, Masulli teaches a functional graph with nodes as points (Masulli [Pg. 2] discusses treating the network as a functional graph in topological space with nodes as points called a clique complex).
Because the combination of Maass and Giusti teaches activity patterns are identified as cliques in the functional graph; and Masulli teaches treating a functional graph of the recurrent neural network as a topological space with nodes as points, accordingly, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate treating a functional graph of the recurrent neural network as a topological space with nodes as points as taught by Masulli, into the combination of Maass and Giusti’s system, with a reasonable expectation of success, to teach wherein identifying decisions in signal transmission activity in the recurrent neural network that is responsive to the input signal comprises treating a functional graph of the recurrent neural network as a topological space with nodes as points, wherein the activity patterns that are identified are cliques in the functional graph of the neural network. This combination would have been motivated by the desire to compute a topological space with the activity patterns identified as cliques (Masulli [Background]).
Claim 44 is rejected under 35 U.S.C. 103 as being unpatentable over Maass, in view of Giusti, as applied in claim 42, and further in view of Yi et al. ("FPGA based spike-time dependent encoder and reservoir design in neuromorphic computing processors", Microprocessors and Microsystems 46 (2016) 175-183, Science Direct) (Year: 2016), hereafter Yi.
Regarding dependent claim 44, the combination of Maass and Giusti teaches the claimed invention as claimed in claim 42, including a recurrent neural network coupled to input signals ([Abstract] discusses readout neurons within a recurrent neural network are coupled to inputs and can learn to extract output from those inputs).
The combination of Maass and Giusti does not explicitly teach wherein the recurrent neural network is trained and executed on an edge device.
However, Yi teaches training and executing a recurrent neural network on an edge device ([Abstract] discusses generating a FPGA for a time dependent encoder, and relies on the dynamical behavior of recurrent neural networks; thus, the RNN is trained and executed on an edge device).
Because the combination of Maass and Giusti teaches a recurrent neural network coupled to input signals; and Yi teaches executing and training a recurrent neural network on an edge device, accordingly, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate training and executing a recurrent neural network on an edge device as taught by Yi, into the combination of Maass and Giusti’s system, with a reasonable expectation of success, to teach wherein the recurrent neural network is trained and executed on an edge device. This combination would have been motivated by the desire to provide recurrent neural networks to edge devices for engineering and scientific applications (Yi [Background]).
Claim 45 is rejected under 35 U.S.C. 103 as being unpatentable over Maass, in view of Giusti, in view of Yi, as applied in claim 44, and further in view of Hinton et al., "Distilling the Knowledge in a Neural Network," CoRR, Submitted on March 9, 2015, arXiv :1503.02531v1, 9 pages), hereafter Hinton.
Hinton was cited in the IDS submitted 06/09/2026.
Regarding dependent claim 45, the combination of Maass, Giusti, and Yi teaches the claimed invention as claimed in claim 44, including the training dataset comprises a plurality of representations of topological structures in patterns of signal transmission activity (Giusti [Pg. 15] discusses collecting all maximal cliques up to five vertices and using it as a training dataset; thus, the cliques represent topological structures in patterns).
The combination of Maass, Giusti, and Yi does not explicitly teach wherein the plurality of representations of topological structures in patterns of signal transmission activity comprises representations of topological structures in patterns of signal transmission activity that occurred in relatively more complex source neural network.
However, Hinton teaches representations of topological structures in patterns of signal transmission activity that occurred in relatively more complex source neural network ([Sec. 1] discusses training a larger, more complex source network, before distilling it into a smaller model that is more suitable; thus, the representations of topological structures in patterns of signal transmission activity came from a relatively more complex source neural network).
Because the combination of Maass, Giusti, and Yi teaches the training dataset comprises a plurality of representations of topological structures in patterns of signal transmission activity; and Hinton teaches representations of topological structures in patterns of signal transmission activity that occurred in relatively more complex source neural network, accordingly, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate representations of topological structures in patterns of signal transmission activity that occurred in relatively more complex source neural network as taught by Hinton, into the combination of Maass, Giusti, and Yi’s system, with a reasonable expectation of success, to teach wherein the plurality of representations of topological structures in patterns of signal transmission activity comprises representations of topological structures in patterns of signal transmission activity that occurred in relatively more complex source neural network. This combination would have been motivated by the desire to make predictions that are less computationally expensive, especially when the models are large neural nets (Hinton [Abstract]).
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Jaeger ("The 'echo state' approach to analysing and training recurrent neural networks - with an Erratum note", Fraunhofer Institute for Autonomous Intelligent Systems, ai.rug) (Year: 2010) ([Abstract] This article investigates what can be gained when RNN states are understood as echo states. Specifically, the article discusses under which conditions echo states arise and describes how RNNs can be trained, exploiting echo states).
Markram (US 2011/0025532 A1, published 02/03/2011) ([Abstract] Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for encoding and decoding information. In one aspect, methods of encoding information in an encoder include the actions of receiving a signal representing information using a collection of discrete digits, converting, by an encoder, the received signal into a time-based code, and outputting the time-based code. The time-based code is divided into time intervals. Each of the time intervals of the time-based code corresponds to a digit in the received signal. Each digit of a first state of the received signal is expressed as a event occurring at a first time within the corresponding time interval of the time-based code. Each digit of a second state of the received signal is expressed as a event occurring at a second time within the corresponding time intervals of the time-based code, the first time is distinguishable from the second time. All of the states of the digits in the received signal are represented by events in the time-based code).
Any inquiry concerning this communication or earlier communications from the examiner should be directed to RILEY S ACOSTA whose telephone number is (571)272-8714. The examiner can normally be reached Monday-Thursday 6am-4pm.
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, Jennifer N Welch can be reached at (571)272-7212. 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.
/RILEY S ACOSTA/Examiner, Art Unit 2143
/JENNIFER N WELCH/Supervisory Patent Examiner, Art Unit 2143
Prosecution Insights