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 .
Claims 1 and 3-17 are presented for examination.
Response to Amendment
Applicant’s amendment has obviated the remaining specification objections. Therefore, those objections are withdrawn.
Information Disclosure Statement
The information disclosure statement (IDS) submitted on September 8, 2026 is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
Claim Rejections - 35 USC § 101
The text of those sections of Title 35, U.S. Code not included in this action can be found in a prior Office action.
Claims 1 and 3-17 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. The analysis of the claims will follow the 2019 Revised Patent Subject Matter Eligibility Guidance, 84 Fed. Reg. 50 (“2019 PEG”).
Claim 1
Step 1: The claim recites a method; therefore, it is directed to the statutory category of processes.
Step 2A Prong 1: The claim recites “setting a total number of nodes … [of] the artificial recurrent neural network;
setting a non-zero number of two or more classes and sub-classes of the nodes … [of] the artificial recurrent neural network, wherein the nodes are discrete computational elements, groups of nodes have a topological relationship to one another and form respective topological elements, wherein the artificial recurrent neural network includes multiple topological elements and activity in a respective of the topological elements indicates a decision by the respective of the topological elements, wherein the decisions by different topological elements are entangled;
setting different structural topological properties of nodes in each class and sub-class, wherein the structural topological properties determine temporal and spatial integration of computations in the topological elements as a function of time as the node combines inputs;
setting different functional properties of nodes in each class and sub-class, wherein the functional properties determine activation, integration, and response functions as a function of time;
setting a number of nodes in each class and sub-class of nodes;
setting a level of structural topological diversity of each node in each class and sub-class of nodes and a level of functional diversity of each node in each class and sub-class of nodes;
setting an orientation of each node in the topology of the … artificial recurrent neural network; and
setting a spatial arrangement of each node in the topology of the … artificial recurrent neural network, … wherein the orientation and spatial arrangement determine which nodes are in communication in the artificial recurrent neural network.” All of these limitations could encompass mentally setting the relevant quantities.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The claim further recites that the method is “computer-implemented … in computer software, firmware, or hardware”, that “setting the spatial arrangement comprises assigning specific locations to the nodes within the hardware of the artificial recurrent neural network,” and that the setting of the claimed parameters occurs “in the hardware of the artificial recurrent neural network”. However, these amount to mere instructions to apply the judicial exception using a generic computer. MPEP § 2106.05(f).
Step 2B: The claim does not contain significantly more than the judicial exception. The analysis at this step is identical to that of step 2A, prong 2. As an ordered whole, the claim is directed to a mentally performable process of selecting properties of a neural network. Nothing in the claim provides significantly more than this. As such, the claim is not patent eligible.
Claim 3
Step 1: A process, as above.
Step 2A Prong 1: The claim recites that “the structural topological properties of nodes include a branching morphology of the nodes and amplitudes and shapes of signals within the nodes, wherein the amplitudes and shapes of signals are set in accordance with a location of a receiving synapse on the branching morphology.” This limitation could encompass mentally setting the amplitudes and shapes of the signals. Setting the structural properties of the nodes remains mentally performable under these further assumptions.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. See claim 1 analysis.
Step 2B: The claim does not contain significantly more than the judicial exception. See claim 1 analysis.
Claim 4
Step 1: A process, as above.
Step 2A Prong 1: The claim recites that “the functional properties of nodes include subthreshold and suprathreshold spiking behavior of the nodes.” Setting the functional properties of the nodes remains mentally performable under these further assumptions.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. See claim 1 analysis.
Step 2B: The claim does not contain significantly more than the judicial exception. See claim 1 analysis.
Claim 5
Step 1: A process, as above.
Step 2A Prong 1: The claim recites that “the number of classes and sub-classes of the nodes in the … artificial recurrent neural network mimics a number of classes and sub-classes of neurons in the target brain tissue.” Setting the number of classes and sub-classes of the nodes remains mentally performable under these further assumptions.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The claim further recites that the number of classes relates to the “hardware of the artificial neural network”. However, this limitation amounts to a mere instruction to apply the judicial exception using a generic computer. MPEP § 2106.05(f).
Step 2B: The claim does not contain significantly more than the judicial exception. The claim further recites that the number of classes relates to the “hardware of the artificial neural network”. However, this limitation amounts to a mere instruction to apply the judicial exception using a generic computer. MPEP § 2106.05(f).
Claim 6
Step 1: A process, as above.
Step 2A Prong 1: The claim recites that “the number of nodes in each class and sub-class of nodes in the artificial recurrent neural network mimics a proportion of the classes and sub-classes of neurons in the target brain tissue.” Setting the number of classes and sub-classes of the nodes remains mentally performable under these further assumptions.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. See claim 1 analysis.
Step 2B: The claim does not contain significantly more than the judicial exception. See claim 1 analysis.
Claim 7
Step 1: A process, as above.
Step 2A Prong 1: The claim recites that “the level of structural diversity and the level of functional diversity of each node in the … artificial recurrent neural network mimics diversity of neurons in the target brain tissue.” Setting the level of structural and functional diversity of the nodes remains mentally performable under these further assumptions.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The claim further recites that the diversity values relate to the “hardware of the artificial neural network”. However, this limitation amounts to a mere instruction to apply the judicial exception using a generic computer. MPEP § 2106.05(f).
Step 2B: The claim does not contain significantly more than the judicial exception. The claim further recites that the diversity values relate to the “hardware of the artificial neural network”. However, this limitation amounts to a mere instruction to apply the judicial exception using a generic computer. MPEP § 2106.05(f).
Claim 8
Step 1: A process, as above.
Step 2A Prong 1: The claim recites that “the orientation of each node in the topology of the artificial recurrent neural network mimics orientation of neurons in the topology of the target brain tissue.” Setting the orientation of the nodes remains mentally performable under these further assumptions.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The claim further recites that the orientations relate to the “hardware of the artificial neural network”. However, this limitation amounts to a mere instruction to apply the judicial exception using a generic computer. MPEP § 2106.05(f).
Step 2B: The claim does not contain significantly more than the judicial exception. The claim further recites that the orientations relate to the “hardware of the artificial neural network”. However, this limitation amounts to a mere instruction to apply the judicial exception using a generic computer. MPEP § 2106.05(f).
Claim 9
Step 1: A process, as above.
Step 2A Prong 1: The claim recites that “the spatial arrangement of each node in the topology of the … artificial recurrent neural network mimics spatial arrangement of the neurons in the topology of the target brain tissue.” Setting the spatial arrangements of the nodes remains mentally performable under these further assumptions.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The claim further recites that the spatial arrangement relates to the “hardware of the artificial neural network”. However, this limitation amounts to a mere instruction to apply the judicial exception using a generic computer. MPEP § 2106.05(f).
Step 2B: The claim does not contain significantly more than the judicial exception. The claim further recites that the spatial arrangement relates to the “hardware of the artificial neural network”. However, this limitation amounts to a mere instruction to apply the judicial exception using a generic computer. MPEP § 2106.05(f).
Claim 10
Step 1: A process, as above.
Step 2A Prong 1: The claim recites that “setting the spatial arrangement comprises setting layers of nodes and/or setting clustering of different classes or subclasses of nodes.” This limitation could encompass mentally determining the architecture of the network’s layers.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. See claim 9 analysis.
Step 2B: The claim does not contain significantly more than the judicial exception. See claim 9 analysis.
Claim 11
Step 1: A process, as above.
Step 2A Prong 1: The claim recites that “setting the spatial arrangement comprises setting nodes for communication between different regions of the … artificial recurrent neural network.” This limitation could encompass mentally designating nodes for use as communicators between regions of the network.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The claim further recites that the regions are of the “hardware of the artificial neural network”. However, this limitation amounts to a mere instruction to apply the judicial exception using a generic computer. MPEP § 2106.05(f).
Step 2B: The claim does not contain significantly more than the judicial exception. The claim further recites that the regions are of the “hardware of the artificial neural network”. However, this limitation amounts to a mere instruction to apply the judicial exception using a generic computer. MPEP § 2106.05(f).
Claim 12
Step 1: A process, as above.
Step 2A Prong 1: The claim recites that “a first of the regions is designated for input of contextual data, a second of the regions is designated for direct input, and a third of the regions is designated for attention input.” The setting of nodes for communication remains mentally performable under these further assumptions.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. See claim 11 analysis.
Step 2B: The claim does not contain significantly more than the judicial exception. See claim 11 analysis.
Claims 13-17
Step 1: The claims recite a non-transitory computer-readable storage medium; therefore, they are directed to the statutory category of articles of manufacture.
Step 2A Prong 1: The claims recite the same judicial exceptions as in claims 1 and 9-12, respectively.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The analysis at this step is identical to that of claims 1 and 9-12, respectively, except insofar as these claims recite “[a]t least one computer-readable storage medium encoded with executable instructions that, when executed by at least one processor, cause the at least one processor to perform operations comprising: constructing nodes of an artificial recurrent neural network that mimics a target brain tissue”. However, this is a mere instruction to apply the judicial exception using a generic computer programmed with a generic class of computer algorithm. MPEP § 2106.05(f).
Step 2B: The claim does not contain significantly more than the judicial exception. See step 2A, prong 2 analysis.
Claim Rejections - 35 USC § 103
Claims 1, 4-10, and 13-15 are rejected under 35 U.S.C. 103 as being unpatentable over Goodman et al., “Brian: A Simulator for Spiking Neural Networks in Python,” in 2 Frontiers in Neuroinformatics 350 (2008) (“Goodman”) in view of Hunzinger et al. (US 20130204814) (“Hunzinger”) and further in view of Arel et al. (US 10325223) (“Arel”), Adachi et al. (US 20160055421) (“Adachi”), and Josue et al. (US 10956812) (“Josue”).
Regarding claim 1, Goodman discloses “[a] method comprising:
constructing an artificial recurrent neural network (see below mapping to Goodman) …, the construction of the artificial recurrent neural network comprising:
setting a total number of nodes in the artificial recurrent neural network (having defined the differential equations, a group P of 4000 neurons is created with these equations [total number of neurons = 4000] – Goodman, p. 4, part B under “Worked Example”), wherein the nodes are discrete computational elements, groups of nodes have a topological relationship to one another and form respective topological elements (network connectivity can be built directly by specifying connectivity per pair of neurons (i, j) [i.e., discrete computational elements], or with all-too-all or random connectivity [topological relationship], where the synaptic weights can be specified by a weight function [topological element = any two neurons plus the connection/weight between them] – Goodman, p. 2, second paragraph under “Features”), wherein the artificial recurrent neural network includes multiple topological elements and activity in a respective of the topological elements indicates a decision by the respective of the topological elements (neuron model is defined by a custom reset function adaptive_threshold_reset which increases the value of voltage Vt by a constant each time a neuron spikes [spike = activity, increasing the value of Vt = decision of the neurons/topological elements]) …;
setting a non-zero number of two or more classes and sub-classes of the nodes in the artificial recurrent neural network (in creating a network structure, two subgroups Pe and Pi of 3200 and 800 nodes, respectively, are created – Goodman, p. 4, part C under “Worked Example” [classes = Pe and Pi, so two classes; one sub-class per class for a total of two sub-classes]);
setting different structural topological properties of nodes in each class and sub-class, wherein the structural topological properties determine temporal and spatial integration of computations in the topological elements as a function of time as the node combines inputs (after creating the excitatory and inhibitory neurons, excitatory and inhibitory connections Ce and Ci are created; Ce specifies that the group Pe should be connected to the variable ge of the group P (the whole group), and similarly for Ci [connectivity of neurons = structural topological properties] – Goodman, pp. 4-5, part C under “Worked Example”; see also P. 1, second paragraph under “Introduction” (describing the neurons as leaky integrate-and-fire, i.e., the connectivity determines the temporal and spatial integration of computations in the elements of the network as the node combines inputs), p. 4, part A under “Worked Example” (showing that the differential equations describing the model are a function of time));
setting different functional properties of nodes in each class and sub-class, wherein the functional properties determine activation, integration, and response functions as a function of time (differential equations [functional properties of nodes] for the model are defined; these equations are used to define an integrate-and-fire neuron with exponential inhibitory and excitatory synapses with different time constants; the differential equation for V defines a leaky integrator [i.e., a functional property determining integration functions] with excitatory [activation] currents ge and inhibitory [response] currents gi- [note also that these differential equations are a function of time] – Goodman, p. 4, part A under “Worked Example”);
setting a number of nodes in each class and sub-class of nodes (in creating a network structure, two subgroups Pe and Pi of 3200 and 800 nodes, respectively, are created – Goodman, p. 4, part C under “Worked Example”); …
setting an orientation of each node in the topology of the … artificial recurrent neural network (see mapping of next limitation); and
setting a spatial arrangement of each node in the topology of the … artificial recurrent neural network, wherein the orientation and spatial arrangement determine which nodes are in communication in the artificial recurrent neural network (having defined the logical network structure, the weight matrix itself is created; each pair of neurons are connected independently at random with probability 0.02 [i.e., the weight matrix determines which neurons are in communication and thereby determines the orientation and spatial arrangement therebetween]; the excitatory synapses have weight 1.62 mV and the inhibitory ones have weight -9 mV – Goodman, p. 6, part D under “Worked Example”).”
Goodman appears not to disclose explicitly the further limitations of the claim. However, Hunzinger discloses “setting a level of structural diversity of each node in each class and sub-class of nodes and a level of functional diversity of each node in each class and sub-class of nodes (in order to represent a value, the activities of a neuron population must be sufficiently diverse such that there exists a set that can estimate the value from those activities; if all neurons have the same dynamics, the activities may be insufficiently diverse to obtain an accurate or precise representation of the value – Hunzinger, paragraph 79; a diverse population of neurons [i.e., neurons with structural diversity] must be used to represent a system; a system requires neurons with different tuning curves (firing rates as a function of input) [functional diversity] – id. at paragraph 83) ….”
Hunzinger and the instant application both relate to neuromorphic systems and are analogous. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Goodman to set a level of functional and structural diversity of the nodes, as disclosed by Hunzinger, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would allow the system to represent a greater range of values than would be available if the neurons were not diverse. See Hunzinger, paragraph 83.
Neither Goodman nor Hunzinger appears to disclose explicitly the further limitations of the claim. However, Arel discloses that “the decisions by different topological elements are entangled (artificial neural networks typically use deep gradient descent and backpropagation to update every node in the artificial neural network for a given output error; this results in a heavily entangled state where every feature element of every node contributes to each output decision – Arel, col. 22, l. 65-col. 23, l. 11) ….”
Arel and the instant application both relate to recurrent neural networks and are analogous. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the combination of Goodman and Hunzinger to entangle the decisions by different topological elements, as disclosed by Arel, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would increase the accuracy of the result by allowing multiple nodes to take part in a decision. See Arel, col. 22, l. 65-col. 23, l. 11.
Neither Goodman, Hunzinger, nor Arel appears to disclose explicitly the further limitations of the claim. However, Adachi discloses “setting … [properties of] the topology of the hardware [including a spatial arrangement] of the … network …, wherein setting the spatial arrangement comprises assigning specific locations to the nodes within the hardware of the … network (iterating locations of nodes comprises selecting a node for movement from a first location to a second location on the hardware graph [network], removing the node from the first location on the hardware graph, and re-placing the node in the second location on the hardware graph – Adachi, claim 2) ….”
Adachi and the instant application both relate to placing computational nodes within hardware and are analogous. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the combination of Goodman, Hunzinger, and Arel to assign locations to nodes within the hardware, as disclosed by Adachi, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would allow neighboring nodes to be placed closer together on the hardware, thereby saving processing power. See Adachi, paragraph 73.
Goodman/Hunzinger/Arel/Adachi appears not to disclose explicitly the further limitations of the claim. However, Josue discloses “mimic[king] a target brain tissue (a neuron is based on neural networks that mimic the way the brain works, such as enabling learning and the ability to detect anomalies, identify similarities and practice “associative memory” – Josue, col. 2, ll. 10-38) ….”
Josue and the instant application both relate to mimicking brains and are analogous. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Goodman/Hunzinger/Arel to mimic the target brain tissue, as disclosed by Josue, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would allow for simultaneously storing and processing information in parallel in a way that is more powerful than the classic von Neumann architecture. See Josue, col. 2, ll. 10-38.
Claim 13 is a computer-readable storage medium claim corresponding to method claim 1 and is rejected for the same reasons as given in the rejection of that claim.
Regarding claim 4, the rejection of claim 1 is incorporated. Goodman further discloses that “the functional properties of nodes include subthreshold and suprathreshold spiking behavior of the nodes (in a thresholding stage, each value of V is compared to a threshold Vt and a list spikes of the indices of the neurons satisfying the condition is returned [i.e., the neuron is added to the spiking list if V is suprathreshold and not if V is subthreshold] – Goodman, p. 5, section F under “Worked Example”).”
Regarding claim 5, the rejection of claim 1 is incorporated. Goodman further discloses “the number of classes and sub-classes of the nodes in the … artificial recurrent neural network [and] a number of classes and sub-classes of neurons (in creating a network structure, two subgroups Pe and Pi of 3200 and 800 nodes, respectively, are created – Goodman, p. 4, part C under “Worked Example” [classes = Pe and Pi, so two classes; one sub-class per class]) ….”
Adachi discloses setting properties of the network in the “hardware of the … network”, as shown in the rejection of claim 1. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the combination of Goodman, Arel, and Hunzinger to assign properties of the network within the hardware, as disclosed by Adachi, for substantially the reasons as given in the rejection of claim 1.
Goodman/Hunzinger/Arel/Adachi appears not to disclose explicitly the further limitations of the claim. However, Josue discloses “mimic[king] … the target brain tissue (a neuron is based on neural networks that mimic the way the brain works, such as enabling learning and the ability to detect anomalies, identify similarities and practice “associative memory” – Josue, col. 2, ll. 10-38).” It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Goodman/Hunzinger/Arel to mimic the target brain tissue, as disclosed by Josue, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would allow for simultaneously storing and processing information in parallel in a way that is more powerful than the classic von Neumann architecture. See Josue, col. 2, ll. 10-38.
Regarding claim 6, the rejection of claim 1 is incorporated. Goodman further discloses “the number of nodes in each class and sub-class of nodes in the artificial recurrent neural network [and] a proportion of the classes and sub-classes of neurons (in creating a network structure, two subgroups Pe and Pi of 3200 and 800 nodes, respectively, are created – Goodman, p. 4, part C under “Worked Example”) ….”
Goodman/Hunzinger/Arel/Adachi appears not to disclose explicitly the further limitations of the claim. However, Josue discloses “mimic[king] … the target brain tissue (a neuron is based on neural networks that mimic the way the brain works, such as enabling learning and the ability to detect anomalies, identify similarities and practice “associative memory” – Josue, col. 2, ll. 10-38).” It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Goodman/Hunzinger/Arel/Adachi to mimic the target brain tissue, as disclosed by Josue, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would allow for simultaneously storing and processing information in parallel in a way that is more powerful than the classic von Neumann architecture. See Josue, col. 2, ll. 10-38.
Regarding claim 7, the rejection of claim 1 is incorporated. Adachi discloses setting properties of the network in the “hardware of the … network”, as shown in the rejection of claim 1. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the combination of Goodman, Arel, and Hunzinger to assign properties of the network within the hardware, as disclosed by Adachi, for substantially the reasons as given in the rejection of claim 1.
Hunzinger further discloses “the level of structural diversity and the level of functional diversity of each node in the artificial recurrent neural network [and] diversity of neurons (in order to represent a value, the activities of a neuron population must be sufficiently diverse such that there exists a set that can estimate the value from those activities; if all neurons have the same dynamics, the activities may be insufficiently diverse to obtain an accurate or precise representation of the value – Hunzinger, paragraph 79; a diverse population of neurons [i.e., neurons with structural diversity] must be used to represent a system; a system requires neurons with different tuning curves (firing rates as a function of input) [functional diversity] – id. at paragraph 83) ….” It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Goodman/Adachi to set a level of functional and structural diversity of the nodes, as disclosed by Hunzinger, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would allow the system to represent a greater range of values than would be available if the neurons were not diverse. See Hunzinger, paragraph 83.
Neither Goodman, Arel, Adachi, nor Hunzinger appears to disclose explicitly the further limitations of the claim. However, Josue discloses “mimic[king] …the target brain tissue (a neuron is based on neural networks that mimic the way the brain works, such as enabling learning and the ability to detect anomalies, identify similarities and practice “associative memory” – Josue, col. 2, ll. 10-38).” It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Goodman/Hunzinger/Arel/Adachi to mimic the target brain tissue, as disclosed by Josue, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would allow for simultaneously storing and processing information in parallel in a way that is more powerful than the classic von Neumann architecture. See Josue, col. 2, ll. 10-38.
Regarding claim 8, the rejection of claim 1 is incorporated. Goodman further discloses “the orientation of each node in the topology of the artificial recurrent neural network [and] orientation of neurons (having defined the logical network structure, the weight matrix itself is created; each pair of neurons are connected independently at random with probability 0.02 [i.e., the weight matrix determines which neurons are in communication and thereby determines the orientation and spatial arrangement therebetween]; the excitatory synapses have weight 1.62 mV and the inhibitory ones have weight -9 mV – Goodman, p. 6, part D under “Worked Example”) ….”
Adachi discloses setting properties of the network in the “hardware of the … network”, as shown in the rejection of claim 1. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the combination of Goodman, Arel, and Hunzinger to assign properties of the network within the hardware, as disclosed by Adachi, for substantially the reasons as given in the rejection of claim 1.
Neither Goodman, Arel, Adachi, nor Hunzinger appears to disclose explicitly the further limitations of the claim. However, Josue discloses “mimic[king] … the topology of the target brain tissue (a neuron is based on neural networks that mimic the way the brain works, such as enabling learning and the ability to detect anomalies, identify similarities and practice “associative memory” – Josue, col. 2, ll. 10-38).” It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Goodman/Hunzinger/Arel/Adachi to mimic the target brain tissue, as disclosed by Josue, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would allow for simultaneously storing and processing information in parallel in a way that is more powerful than the classic von Neumann architecture. See Josue, col. 2, ll. 10-38.
Regarding claim 9, the rejection of claim 1 is incorporated. Goodman further discloses “the spatial arrangement of each node in the topology of the artificial recurrent neural network [and the] spatial arrangement of the neurons (having defined the logical network structure, the weight matrix itself is created; each pair of neurons are connected independently at random with probability 0.02 [i.e., the weight matrix determines which neurons are in communication and thereby determines the orientation and spatial arrangement therebetween]; the excitatory synapses have weight 1.62 mV and the inhibitory ones have weight -9 mV – Goodman, p. 6, part D under “Worked Example”) ….”
Adachi discloses setting properties of the network in the “hardware of the … network”, as shown in the rejection of claim 1. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the combination of Goodman, Arel, and Hunzinger to assign properties of the network within the hardware, as disclosed by Adachi, for substantially the reasons as given in the rejection of claim 1.
Goodman/Hunzinger/Arel/Adachi appears not to disclose explicitly the further limitations of the claim. However, Josue discloses “mimic[king] … the topology of the target brain tissue (a neuron is based on neural networks that mimic the way the brain works, such as enabling learning and the ability to detect anomalies, identify similarities and practice “associative memory” – Josue, col. 2, ll. 10-38).” It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Goodman/Hunzinger/Arel/Adachi to mimic the target brain tissue, as disclosed by Josue, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would allow for simultaneously storing and processing information in parallel in a way that is more powerful than the classic von Neumann architecture. See Josue, col. 2, ll. 10-38.
Claim 14 is a computer-readable storage medium claim corresponding to method claim 9 and is rejected for the same reasons as given in the rejection of that claim.
Regarding claim 10, the rejection of claim 9 is incorporated. Goodman further discloses that “setting the spatial arrangement comprises setting layers of nodes and/or setting clustering of different classes or subclasses of nodes (in creating a network structure, two subgroups [clusters] Pe and Pi of 3200 and 800 nodes, respectively, are created, where “Pe” is the excitatory subgroup [cluster] and “Pi” is the inhibitory subgroup [cluster] – Goodman, pp. 4-6, part C under “Worked Example” [classes = Pe and Pi, so two classes; one sub-class per class]).”
Claim 15 is a computer-readable storage medium claim corresponding to method claim 10 and is rejected for the same reasons as given in the rejection of that claim.
Claim 3 is rejected under 35 U.S.C. 103 as being unpatentable over Goodman in view of Hunzinger, Adachi, Josue, and Arel and further in view of Koelmans et al. (US 20190065929) (“Koelmans”).
Regarding claim 3, Goodman/Hunzinger/Arel/Adachi/Josue appears not to disclose explicitly the further limitations of the claim. However, Koelmans discloses that “the structural topological properties of nodes include a branching morphology of the nodes and amplitudes and shapes of signals within the nodes, wherein the amplitudes and shapes of signals are set in accordance with a location of a receiving synapse on the branching morphology (short term potentiation/long term potentiation rules may depend on variables such as the shape and amplitude of the spikes [amplitudes and shapes of signals within the nodes], as well as properties of post-neuron circuits [receiving synapses] – Koelmans, paragraph 30; system may form part of an array of interconnected neuron circuits, with each neuron circuit being connected to other neuron circuits via respective synapses [connections among circuits = branching morphology] – id. at paragraph 23; independent control of volatile and non-volatile properties may be provided by modulating aspects of the control signal applied to a terminal, which include parameters such as amplitude, shape, and other controlled properties of the electrical signal that could differ between different terminals [i.e., the shape and amplitude of the signal depend on the terminal to which the signal is sent] – id. at paragraph 54).”
Koelmans and the instant application both relate to neuromorphic computing architectures and are analogous. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Goodman/Hunzinger/Arel/Adachi/Josue to set amplitudes and shapes of signals in accordance with destination synapses, as disclosed by Koelmans, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would allow for a more fine-grained control of the system than would be possible if the amplitudes and shapes of all signals were independent of destination. See Koelmans, paragraph 54.
Claims 11-12 and 16-17 are rejected under 35 U.S.C. 103 as being unpatentable over Goodman in view of Hunzinger, Adachi, Josue, and Arel and further in view of Kim et al. (US 20200311207) (“Kim”).
Regarding claim 11, Adachi discloses setting properties of the network in the “hardware of the … network”, as shown in the rejection of claim 1. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the combination of Goodman, Arel, Josue, and Hunzinger to assign properties of the network within the hardware, as disclosed by Adachi, for substantially the reasons as given in the rejection of claim 1.
Neither Goodman, Arel, Adachi, Josue, nor Hunzinger appears to disclose explicitly the further limitations of the claim. However, Kim discloses that “setting the spatial arrangement comprises setting nodes for communication between different regions of the artificial recurrent neural network (neural network system can include vector component, attention component, similarity component, and labeling component [components = regions] – Kim, paragraph 48; neural network system comprises a plurality of interconnected nodes – id. at paragraph 58; see also Fig. 2 (showing that vector component, attention component, similarity component, and labeling component are in communication with one another)).”
Kim and the instant application both relate to neural networks and are analogous. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Goodman/Hunzinger/Arel/Adachi/Josue to set up regions of the neural network in communication with each other, as disclosed by Kim, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would ensure that the network is capable of executing multiple functionalities, thereby increasing its power. See Kim, paragraph 48.
Claim 16 is a computer-readable storage medium claim corresponding to method claim 11 and is rejected for the same reasons as given in the rejection of that claim.
Regarding claim 12, neither Goodman, Arel, Adachi, Josue, nor Hunzinger appears to disclose explicitly the further limitations of the claim. However, Kim discloses that “a first of the regions is designated for input of contextual data, a second of the regions is designated for direct input, and a third of the regions is designated for attention input (neural network system can include vector component [direct input region], attention component [attention input region], similarity component [contextual data region], and labeling component – Kim, paragraph 48; vector component can be used to extract information from input data [i.e., it is designated for direct input] – id. at paragraph 50; similarity component explicitly adds similarity information to the text segmentation neural network system; similarity information can be based on a comparison between combinations of the context vectors and the target vectors to determine differences [i.e., it is for input of, inter alia, contextual data] – id. at paragraph 55).” It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Goodman/Hunzinger/Arel/Adachi/Josue to include a contextual data region, a direct input region, and an attention region among the regions of the network, as disclosed by Kim, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would ensure that the network is capable of executing multiple functionalities, thereby increasing its power. See Kim, paragraph 48.
Claim 17 is a computer-readable storage medium claim corresponding to method claim 12 and is rejected for the same reasons as given in the rejection of that claim.
Response to Arguments
Applicant's arguments filed September 8, 2026 (“Remarks”) have been fully considered but they are, except insofar as rendered moot by the introduction of a new ground of rejection, not persuasive.
Applicant argues that the amended claims are eligible under 35 USC § 101 because (a) setting quantities in a neural network is not mentally performable; and (b) the claims effect an improvement to a technological field by mimicking biological brain tissue, that this alleged improvement to technology is disclosed by the specification, and that the claims are directed to a non-conventional architecture for hardware that results in improved data processing. Remarks at 7-8. However, regarding (a), insofar as deciding what those quantities will be is mentally performable, the relevant limitations still recite an abstract idea. Insofar as the claim now generically recites that those settings are to be placed in hardware of a neural network, that recitation is a mere instruction to apply the judicial exception on a generic computer. MPEP § 2106.05(f). Regarding (b), even assuming arguendo that the specification discloses an improvement to technology, which Examiner does not concede, that alleged improvement is not reflected in the additional elements of the claims themselves. The judicial exception cannot provide the inventive concept, MPEP § 2106.05(I), and the only additional elements recite generic hardware implementation of the judicial exception of setting various neural network quantities. Since the majority of the limitations on which Applicant is implicitly relying form part of the judicial exception, it is irrelevant whether they are conventional, or even disclosed by the prior art, or not.
Applicant then argues that the amended claims are distinguishable over the prior art of record because Goodman’s neurons allegedly lack orientations and spatial arrangements in hardware. Remarks at 9. This argument is moot by virtue of the use of Adachi to teach the spatial orientation of neurons in hardware.
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.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to RYAN C VAUGHN whose telephone number is (571)272-4849. The examiner can normally be reached M-R 7:00a-5:00p ET.
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, Kamran Afshar, can be reached at 571-272-7796. 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.
/RYAN C VAUGHN/ Primary Examiner, Art Unit 2125
Prosecution Insights