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 .
DETAILED ACTION
1. This Office Action is in response to the application filed on 05/21/2024.
Claims 1-9 are pending.
Priority
2. Receipt is acknowledged of papers submitted under 35 U.S.C. 119(a)-(d), which papers have been placed of record in the file.
Information Disclosure Statement
3. The information disclosure statement (IDS) filed on 05/21/2024 complies with the provisions of M.P.E.P. 609. The examiner has considered it.
Claim Rejections - 35 USC § 101
4. 35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
5. Claims 1-9 are rejected under 35 U.S.C. 101 because the claimed invention is directed to non-statutory subject matter.
At Step 1:
Independent claims 1, 8 and 9 are directed to a "system", a “device” and a “method” and thus directed to a statutory category
At Step 2A, Prong One:
The claim recites the following limitations directed to an abstract idea:
• " train the spiking neural network using an evaluation function that indicates a
better evaluation as a probability that an individual neuron model fires decreases" as drafted this recites a mentally performable process as an evaluation or judgement. This is also consistent with the specification as in Fig. 13 and page 56 and paragraph 2 where one can mentally visualizing train the spiking neural network using an evaluation function that indicates a better evaluation as a probability that an individual neuron model fires decreases .
At Step 2A, Prong Two:
• The claim recites no additional elements. At most one might consider that a " at least one memory …" as claimed might be considered to represent a computer-implemented system and method consistent with Fig. 1 even though the claim does not recite any computer. At most this would be a high-level recitation of a generic computer components and represents mere instructions to apply the abstract idea on a computer as in MPEP 2106.05(f), which does not provide integration into a practical application.
• Viewing the additional limitations together and the claim as a whole, nothing provides integration into a practical application.
At Step 2B:
• The conclusions for the mere implementation using a computer are carried over and does not provide significantly more.
• Looking at the claim as a whole does not change this conclusion and the claim is ineligible.
Dependent Claims 2-7
The limitations as recited in dependent claim 2 recites, “train the spiking neural network using the evaluation function that indicates a better evaluation as a value of a time integral relating to a membrane potential of the neuron model decreases which further describes the concepts performed in the human mind including an observation, evaluation, judgment, and opinion, in step 2A prong one.
The limitations as recited in dependent claim 3 recites, “train the spiking neural network using the evaluation function that includes a sub-expression representing, for a neuron model that has fired, a total value obtained by dividing an integral of a difference between a membrane potential and a set value during a time the membrane potential of the neuron model is greater than or equal to the set value, being a smaller value than a threshold membrane potential, and less than or equal to the threshold membrane potential, by a difference between the threshold membrane potential and the set value”, which further describes the concepts performed in the human mind including an observation, evaluation, judgment, and opinion and mathematical relationships, in step 2A prong one.
The limitations as recited in dependent claim 4 recites, “train the spiking neural network using a learning method that uses the evaluation function that takes a limit that brings the set value close to the threshold membrane potential, and takes a derivative of the evaluation function”, which further describes the concepts performed in the human mind including an observation, evaluation, judgment, and opinion and mathematical relationships, in step 2A prong one.
The limitations as recited in dependent claim 5 recites, “train the spiking neural network using the evaluation function obtained by representing a membrane potential of the neuron model using a weighting coefficient of a spike that has been input to the neuron model, and a firing time of a neuron model that has output a spike to the neuron model”, which further describes the concepts performed in the human mind including an observation, evaluation, judgment, and opinion and mathematical relationships, in step 2A prong one.
The limitations as recited in dependent claim 6 recites, “perform differentiation of the evaluation function by treating a time interval from a time at which the membrane potential has reached the set value to a firing time of the neuron model, as a fixed time interval”, which further describes the concepts performed in the human mind including an observation, evaluation, judgment, and opinion and mathematical relationships, in step 2A prong one.
The limitations as recited in dependent claim 7 recites, “train the spiking neural network using the evaluation function that, for a neuron model that has fired, indicates a better evaluation as a total value of weighting coefficients of spikes that have been input to the neuron model decreases”, which further describes the concepts performed in the human mind including an observation, evaluation, judgment, and opinion and mathematical relationships, in step 2A prong one.
Examiner’s Note
6. Spiking Neural Networks (According to Google): “Spiking Neural Networks (SNNs) are a brain-inspired type of artificial neural network. Unlike traditional networks that evaluate inputs using continuous values, SNNs communicate using discrete electrical events called "spikes." They only process information or consume energy when a firing threshold is met, making them incredibly energy-efficient. How SNNs Work: (1) Biological Inspiration: Unlike standard multi-layer perceptrons, SNNs directly mimic the information processing of biological neurons in the human brain. (2) The "Integrate-and-Fire" Mechanism: An SNN neuron accumulates incoming electrical signals. Once its internal membrane potential reaches a specific threshold, it "fires" or emits a spike that is transmitted to the next layer of neurons. (3) Temporal and Pulse Coding: Instead of relying on the magnitude of values like traditional AI, SNNs often use temporal coding, where the precise timing and frequency of the spikes carry the meaningful information. SNNs vs. Traditional ANNs: (1) Sparsity & Efficiency: In a standard Convolutional Neural Network (CNN) or Recurrent Neural Network (RNN), every neuron computes an output in every cycle, which consumes a massive amount of power. SNNs are inherently sparse—they remain quiet unless an event triggers them, saving significant energy. (2) Learning Rules: Because discrete spikes are non-differentiable, SNNs often rely on biologically plausible learning rules like Spike-Timing-Dependent Plasticity (STDP) or surrogate gradients rather than traditional backpropagation. (3) Hardware Compatibility: SNNs are designed to operate natively on specialized, low-power neuromorphic hardware (like Intel Loihi). This hardware uses asynchronous, parallel processing to execute computations efficiently.”
What does this phrase, " train the spiking neural network using an evaluation function that indicates a better evaluation as a probability that an individual neuron model fires decreases”, mean? (According to Google): “This phrase describes an optimization or learning rule where the network is rewarded for reducing how often neurons fire. In tasks like probabilistic inference, minimizing firing rates forces the model to encode information more efficiently, relying only on the most necessary spikes to represent a state or solution. Here is the breakdown of exactly how this works: (1) Spiking Neural Network (SNN): Unlike standard deep learning models that pass continuous values, SNNs compute by sending binary "spikes" (either firing a 1 or staying silent at 0). (2) Evaluation (or Cost) Function: A mathematical equation used to score the network’s performance. A "better evaluation" means a higher reward or a lower error. (3) Probability of Firing Decreases: The likelihood or frequency of a neuron model emitting a spike goes down. (4) The Meaning: The algorithm iteratively adjusts the network. Every time it changes weights to reduce the overall firing rate of its neurons, the evaluation score goes up.”.
Cao et al, US 20200272815, [Cao: Paragraph 14 (“as part of a training process by which the spiking neural network learns to detect for indicia of a given data type (or “data class”). Alternatively or in addition, feedback signaling may be provided as part of a process to detect whether some data under test is an instance of a data type which the spiking neural network was previously trained to recognize”)] [Cao: Paragraph 17 (“The term “reference data” refers to data which is to be used as a basis for training of a spiking neural network to recognize one or more data types. “Reference signaling” refers herein to one or more signals (e.g., including one or more spike trains) which represent an encoded version of reference data—e.g., wherein such reference signaling is provided to, or is generated by, a spiking neural network to facilitate training thereof.”)] [Cao: Paragraph 26 (“In a spiking neural network, the artificial neuron “fires” (e.g., produces an output spike), when its membrane potential crosses a firing threshold. Thus, the effect of inputs on a spiking neural network neuron operate to increase or decrease its internal membrane potential, making the neuron more or less likely to fire. Further, in a spiking neural network, input connections may be stimulatory or inhibitory. A neuron's membrane potential may also be affected by changes in the neuron's own internal state (“leakage”)”)] [Cao: Paragraph 38 (“applying a trained spiking neural network to the encoded image data, wherein one or more nodes of the trained spiking neural network each receive a respective spiked input signal based on the encoded image data. Although some embodiments are not limited in this regard, method 200 may additionally or alternatively include operations (not shown)—e.g. performed prior to the applying at 230—by which the spiked neural network is trained. For example, such operations may include training the spiking neural network based on one or more reference images which each represent a low-noise representation of a respective object which the spiking neural network is to be configured to recognize. Such training of the spiking neural network may comprise a forward training stage and a feedback training stage which are each based on the one or more reference images”)].
Wozniak et al, US 11,727,252, [Wozniak: Column 1, lines 10-18 (“Neural networks are a computational model used in artificial intelligence systems. Neural networks are based on multiple artificial neurons. Each artificial neuron is connected with one or more other neurons, and links can enhance or inhibit the activation state of adjoining neurons. Artificial neurons such as neurons of a spiking neural network (SNN) are provided with a firing threshold that must be exceeded by a membrane potential of the neurons in order to generate a spike.”)] [Wozniak: Column 5, lines 60-67 through column 6, lines 1-11 (“Neural networks are a computational model used in artificial intelligence systems. Neural networks are based on multiple artificial neurons. Each artificial neuron is connected with one or more other neurons, and links can enhance or inhibit the activation state of adjoining neurons. Artificial neurons such as neurons of a spiking neural network (SNN) are provided with a firing threshold that must be exceeded by a membrane potential of the neurons in order to generate a spike.”)].
Seller, US 20260148047, [Seller: Abstract and paragraphs 1, 6 and 9 (“an artificial spiking neural network building block, a spiking neural network, a method for training a spiking neural network”)] [Seller: Paragraphs 38 and 43 (“the whole duration of the refractory period (considered equal to 1 millisecond, for example). This can be achieved by keeping track or storing the time of firing and making sure that no other firing event happens in the amount of time specified as the refractory period”)].
UIAmeen Sheik et al, US 20230385617, [Sheik: Paragraphs 2 and 7 (“In order to improve a training efficiency of a spiking neural network, the present invention achieves the purpose in the following: A signal processing method for neurons in a spiking neural network, wherein the spiking neural network comprises a plurality of layers, each of the layers comprises a plurality of neurons, and the signal processing method comprises following steps: a receiving step: at least one neuron configured to receive at least one path of input spike train; an accumulation step: performing weighted summation based on the at least one path of input spike train to obtain a membrane potential; and an activation step: when the membrane potential exceeds a threshold value, determining an amplitude of a spike fired by the at least one neuron based on a ratio of the membrane potential to the threshold value”)] [Sheik: Paragraphs 8-9 and 16 (“A training method of a spiking neural network, wherein the spiking neural network comprises a plurality of layers, and each of the layers comprises a plurality of neurons, comprising: when the neurons process signals in a network training, following steps are included: a receiving step: at least one neuron configured to receive at least one path of input spike train; an accumulation step: performing weighted summation based on the at least one path of input spike train to obtain a membrane potential; and an activation step: when the membrane potential exceeds a threshold value, determining an amplitude of a spike fired by the at least one neuron based on a ratio of the membrane potential to the threshold value; wherein a total loss of the spiking neural network comprises a first loss and a second loss, the first loss reflects a gap between an expected output of the spiking neural network and an actual output of the spiking neural network, and the second loss reflects an activity or an activity level of the at least one neuron”)] [Sheik: Paragraph 57 (“That is, at a single simulation time step, the amplitude of the fired spike is related to the ratio of the membrane potential to the threshold value. “s′=1” here means a spike with unit amplitude (i.e., unit spike). That is, the above formula discloses that at a single simulation time step, the ratio of the amplitude of the fired spike to the unit spike amplitude is equal to the rounded down value of the ratio of the membrane potential to the threshold value”)].
Chakrabartty et al, US 20230021621, [Chakrabartty: Paragraph 6 (“backpropagation remains the tool of choice for training spiking neural networks”)] [Chakrabartty: Paragraphs 7-8 (“For example, a spike, or high firing rate, in response to a positive pattern and silence, or low firing rate, otherwise. Examples include supervised synaptic learning rules, such as the tempotron implementing temporal credit assignments according to elicited output responses and algorithms using teaching signals to drive outputs in the decision layer”)] [Chakrabartty: Paragraph 54 (“the trained neural network may enforce sparsity in a network's spiking activity.”)] [Chakrabartty: Paragraph 81 (“In some embodiments, a modulation function may provide another degree of freedom that may be varied to model different transient firing statistics based on local and/or global variables. For example, the modulation function may be varied based on local variables, such as membrane potentials or local firing rates, to reproduce single-neuron response characteristics like tonic spiking, bursting, spike-frequency adaptation, etc., or based on global properties like the state of convergence of the network to yield different population-level dynamics”)] [Chakrabartty: Paragraph 99 (“IG. 6D pictorially depicts how the sparsest solution is achieved through firing rate minimization for a differential network M=2. In this example, the matrix Q is symmetric and the solution can be visualized using energy contours”)] [Chakrabartty: Paragraphs 100, 111, 129 and 141 (“Similarly, the negative network may move toward the path NO, given by the vector sum of paths NC and NB. This may minimize the overall firing rate of the network and drives the membrane potentials of each differential pair towards zero, while simultaneously ensuring that the linear constraint in (36) is always satisfied”)].
Claim Rejections - 35 USC § 102
7. 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 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.
8. The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale or otherwise available to the public before the effective filing date of the claimed invention.
9. Claims 1-9 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Jason K. Eshraghian et al ("Training Spiking Neural Networks Using Lessons From Deep Learning", Jan 14, 2022 (Year: 2022)).
Claim 1:
Eshraghian suggests a neural network system comprising: a time-based spiking neural network configured using a neuron model based on an integrate-and-fire model [Eshraghian: Title and Abstract (“Spiking Neural Networks”)]; at least one memory configured to store instructions; and at least one processor configured to execute the instructions to: train the spiking neural network using an evaluation function that indicates a better evaluation as a probability that an individual neuron model fires decreases [Eshraghian: Title and Abstract (“Training Spiking Neural Networks Using Lessons From Deep Learning”)] [Eshraghian: Page 4, section 2.1 “Spiking Nerons”)] [Eshraghian: Page 6 ,i.e., ‘evaluation function’)] [Eshraghian: Page 7, section 3 and section 3.1 (1. Rate coding converts input intensity into a firing rate or spike count”)] [Eshraghian: Page 8 (“An alternative method counts the spikes over a very short time interval t. For a small enough t, the spike count can be constrained to either 0 or 1, limiting the total number of possible outcomes to only two. By repeating this experiment multiple times, the average number of spikes (over trials) occurring within t can be found. This average must be equal to or less than 1, interpreted as the observed probability that a neuron will fire within the brief time interval. To convert it into a time-dependent firing rate, the trial-average is divided by the duration of the interval. This probabilistic interpretation of the rate code can be distributed across multiple neurons, where counting up the spikes from a collection of neurons advocates for a population code”)].
Claim 2:
Eshraghian suggests wherein the at least one processor is configured to train the spiking neural network using the evaluation function that indicates a better evaluation as a value of a time integral relating to a membrane potential of the neuron model decreases [Eshraghian: Page 7 (“Sout[t] 2 f0; 1g is the output spike generated by the neuron, where if activated (Sout = 1), the reset term subtracts the threshold from the membrane potential”)].
Claim 3:
Eshraghian suggests wherein the at least one processor is configured to train the spiking neural network using the evaluation function that includes a sub-expression representing, for a neuron model that has fired, a total value obtained by dividing an integral of a difference between a membrane potential and a set value during a time the membrane potential of the neuron model is greater than or equal to the set value, being a smaller value than a threshold membrane potential, and less than or equal to the threshold membrane potential, by a difference between the threshold
membrane potential and the set value [Eshraghian: Page 7 (“Sout[t] 2 f0; 1g is the output spike generated by the neuron, where if activated (Sout = 1), the reset term subtracts the threshold from the membrane potential”)].
Claim 4:
Eshraghian suggests wherein the at least one processor is configured to train the spiking neural network using a learning method that uses the evaluation function that takes a limit that brings the set value close to the threshold membrane potential, and takes a derivative of the evaluation function [Eshraghian: Page 7 (“Sout[t] 2 f0; 1g is the output spike generated by the neuron, where if activated (Sout = 1), the reset term subtracts the threshold from the membrane potential”)].
Claim 5:
Eshraghian suggests wherein the at least one processor is configured to train the spiking neural network using the evaluation function obtained by representing a membrane potential of the neuron model using a weighting coefficient of a spike that has been input to the neuron model, and a firing time of a neuron model that has output a spike to the neuron model [Eshraghian: Page 7 (“Sout[t] 2 f0; 1g is the output spike generated by the neuron, where if activated (Sout = 1), the reset term subtracts the threshold from the membrane potential”)].
Claim 6:
Eshraghian suggests wherein the at least one processor is configured to perform differentiation of the evaluation function by treating a time interval from a time at which the membrane potential has reached the set value to a firing time of the neuron model, as a fixed time interval [Eshraghian: Page 7 (“Sout[t] 2 f0; 1g is the output spike generated by the neuron, where if activated (Sout = 1), the reset term subtracts the threshold from the membrane potential”)].
Claim 7:
Eshraghian suggests wherein the at least one processor is configured to train the spiking neural network using the evaluation function that, for a neuron model that has fired, indicates a better evaluation as a total value of weighting coefficients of spikes that have been input to the neuron model decreases [Eshraghian: Page 7 (“Sout[t] 2 f0; 1g is the output spike generated by the neuron, where if activated (Sout = 1), the reset term subtracts the threshold from the membrane potential” and “In deep learning, the weighting factor of an input is typically a learnable parameter. Relaxing the physically viable assumptions made thus far, the coefficient of input current in Equation (3), (1 ), is subsumed into a learnable weight W, and the simplification Iin[t] = WX[t] is made to decouple the effect of on the input X[t].”)].
Claim 8:
Claim 8 is essentially the same as claim 1 except that it sets forth the claimed invention as a device rather than a system and rejected under the same reasons as applied above.
Claim 9:
Claim 9 is essentially the same as claim 1 except that it sets forth the claimed invention as a method rather than a system and rejected under the same reasons as applied above.
10. Any inquiry concerning this communication or earlier communications from the examiner should be directed to [Hung D. Le], whose telephone number is [571-270-1404]. The examiner can normally be communicated on [Monday to Friday: 9:00 A.M. to 5:00 P.M.].
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Apu Mofiz can be reached on [571-272-4080]. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of an application may be obtained from the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see http://pair-direct.uspto.gov. Should you have questions on access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative or access to the automated information system, contact [800-786-9199 (IN USA OR CANADA) or 571-272-1000].
Hung Le
07/22/2026
/HUNG D LE/
Primary Examiner
Prosecution Insights