DETAILED ACTION
This action is made FINAL in response to the amendments filed on 8/13/2026.
Claim Rejections - 35 USC § 101
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.
Claims 1 – 14 and 18 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step One
The claims are directed to a method (claims 1 – 12 and 18), and apparatus (claim 13), and a non-transitory computer readable medium (claim 14). Thus, each of the claims falls within one of the four statutory categories (i.e., process, machine, manufacture, or composition of matter).
As to claim 1,
Step 2A, Prong One
The claim recites in part:
selecting a type included in data to be input to the block by processing based on a genetic algorithm in learning using the learning data, and
generating the model by using data corresponding to a combination of types selected among the plurality of types as an input from the input layer to the block.
As drafted and under its broadest reasonable interpretation, these limitation covers performance of the limitation in the mind (including an observation, evaluation, judgment, opinion) or with the aid of pencil and paper but for the recitation of generic computer components. For example, a human can select a category (type) of data based on the results of a genetic algorithm and generate rules and patterns (model) based on the different types of data.
Accordingly, at Step 2A, Prong One, the claim is directed to an abstract idea.
Step 2A, Prong Two
The judicial exception is not integrated into a practical application. In particular, the claim recites the additional elements of:
acquiring learning data used for learning of a model comprising a neural network having an input layer, at least one intermediate layer, and an output layer, the model having at least one block to which an output from the input layer is input, the learning data including a plurality of types of information
which amounts to extra-solution activity of gathering data for use in the claimed process. As described in MPEP 2106.05(g), limitations that amount to merely adding insignificant extra-solution activity to a judicial exception do not amount to significantly more than the exception itself, and cannot integrate a judicial exception into a practical application.
The claim further recites:
executing inference processing on a basis of output data output by the model by using data in which a masking type that is a type to be partially masked among the combination of types is masked as an input to the block of the model.
which is recited at a high-level of generality with no detail of the execution of the inference processing and amounts to no more than adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea. Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea (See MPEP 2106.05(f))
Accordingly, at Step 2A, Prong Two, the additional elements individually or in combination do no integrate the judicial exception into a practical application.
Step 2B
In accordance with Step 2B, the claim does not include additional elements that are sufficient to amount to significantly more that the judicial exception. As discussed above, the additional elements of:
acquiring learning data used for learning of a model comprising a neural network having an input layer, at least one intermediate layer, and an output layer, the model having at least one block to which an output from the input layer is input, the learning data including a plurality of types of information
are recited at a high level of generality and amounts to extra-solution activity of receiving data i.e. pre-solution activity of gathering data for use in the claimed process. The courts have found limitations directed to obtaining information electronically, recited at a high level of generality, to be well-understood, routine, and conventional (see MPEP 2106.05(d)(II), “receiving or transmitting data over a network”, "electronic record keeping," and "storing and retrieving information in memory").
The claim further recites:
executing inference processing on a basis of output data output by the model by using data in which a masking type that is a type to be partially masked among the combination of types is masked as an input to the block of the model.
which is recited at a high-level of generality with no detail of the execution of the inference processing and amounts to no more than adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea. Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea (See MPEP 2106.05(f))
Accordingly, at Step 2B the additional elements individually or in combination do not amount to significantly more than the judicial exception.
As to claim 2,
Step 2A, Prong One
The claim recites in part:
determining the combination of types partially used as an input to the block at a time of inference using the model.
As drafted and under its broadest reasonable interpretation, these limitation covers performance of the limitation in the mind (including an observation, evaluation, judgment, opinion) or with the aid of pencil and paper but for the recitation of generic computer components. For example, a human can determine the different combinations of inputs based on past rules and patterns.
Accordingly, at Step 2A, Prong One, the claim is directed to an abstract idea.
Step 2A, Prong Two
The claim does not include additional elements that integrate the judicial exception into a practical application.
Step 2B
The claim does not include additional elements that are sufficient to amount to “significantly more” to the judicial exception.
As to claim 3,
Step 2A, Prong One
The claim recites in part:
determining a type to be masked at a time of inference using the model among the combination of types.
As drafted and under its broadest reasonable interpretation, these limitation covers performance of the limitation in the mind (including an observation, evaluation, judgment, opinion) or with the aid of pencil and paper but for the recitation of generic computer components. For example, a human can determine the different the types of input data to be masked based on the past rules and patterns.
Accordingly, at Step 2A, Prong One, the claim is directed to an abstract idea.
Step 2A, Prong Two
The claim does not include additional elements that integrate the judicial exception into a practical application.
Step 2B
The claim does not include additional elements that are sufficient to amount to “significantly more” to the judicial exception.
As to claim 4,
Step 2A, Prong One
The claim recites in part:
generating the model in which the combination of types included in the data input from the input layer to the block is determined by combination optimization based on the genetic algorithm
As drafted and under its broadest reasonable interpretation, these limitation covers performance of the limitation in the mind (including an observation, evaluation, judgment, opinion) or with the aid of pencil and paper but for the recitation of generic computer components. For example, a human can determine the optimal combinations of inputs based on past rules and patterns
Accordingly, at Step 2A, Prong One, the claim is directed to an abstract idea.
Step 2A, Prong Two
The claim does not include additional elements that integrate the judicial exception into a practical application.
Step 2B
The claim does not include additional elements that are sufficient to amount to “significantly more” to the judicial exception.
As to claim 5,
Step 2A, Prong One
The claim recites in part:
generating the model in which the combination of types included in the data input from the input layer to the block is determined by a search based on the genetic algorithm.
As drafted and under its broadest reasonable interpretation, these limitation covers performance of the limitation in the mind (including an observation, evaluation, judgment, opinion) or with the aid of pencil and paper but for the recitation of generic computer components. For example, a human can determine the where and what to search based on the output of the genetic algorithm.
Accordingly, at Step 2A, Prong One, the claim is directed to an abstract idea.
Step 2A, Prong Two
The claim does not include additional elements that integrate the judicial exception into a practical application.
Step 2B
The claim does not include additional elements that are sufficient to amount to “significantly more” to the judicial exception.
As to claim 6,
Step 2A, Prong One
The claim recites in part:
determining the masking type at a time of learning the model by measuring accuracy of the model while changing a masking type candidate that is a type among the combination of types, and determining, as the masking type, the masking type candidate for which the accuracy is best
As drafted and under its broadest reasonable interpretation, these limitation covers performance of the limitation in the mind (including an observation, evaluation, judgment, opinion) or with the aid of pencil and paper but for the recitation of generic computer components. For example, a human can determine the masking type by ranking the masking types based on accuracy and selecting the top ranked (or best) masking type.
Accordingly, at Step 2A, Prong One, the claim is directed to an abstract idea.
Step 2A, Prong Two
The claim does not include additional elements that integrate the judicial exception into a practical application.
Step 2B
The claim does not include additional elements that are sufficient to amount to “significantly more” to the judicial exception.
As to claim 7,
Step 2A, Prong One
The claim recites in part:
executing the inference processing on a basis of output data output by the model by using input data corresponding to the combination of types determined as an input to the block of the mode
As drafted and under its broadest reasonable interpretation, these limitation covers performance of the limitation in the mind (including an observation, evaluation, judgment, opinion) or with the aid of pencil and paper but for the recitation of generic computer components. For example, a human can determine combinations of new input data based on the rules and patterns of past input data.
Accordingly, at Step 2A, Prong One, the claim is directed to an abstract idea.
Step 2A, Prong Two
The claim does not include additional elements that integrate the judicial exception into a practical application.
Step 2B
The claim does not include additional elements that are sufficient to amount to “significantly more” to the judicial exception.
As to claim 8,
Step 2A, Prong One
The claim recites in part:
executing the inference processing on a basis of output data output by the model by using data corresponding to only a part of the combination of types determined as an input to the block of the model.
As drafted and under its broadest reasonable interpretation, these limitation covers performance of the limitation in the mind (including an observation, evaluation, judgment, opinion) or with the aid of pencil and paper but for the recitation of generic computer components. For example, a human can determine combinations of new input data based on the rules and patterns of past input data.
Accordingly, at Step 2A, Prong One, the claim is directed to an abstract idea.
Step 2A, Prong Two
The claim does not include additional elements that integrate the judicial exception into a practical application.
Step 2B
The claim does not include additional elements that are sufficient to amount to “significantly more” to the judicial exception.
As to claim 9,
Step 2A, Prong One
The claim recites in part:
determining the masking type by a search based on the genetic algorithm performed in consideration of a type to be masked.
As drafted and under its broadest reasonable interpretation, these limitation covers performance of the limitation in the mind (including an observation, evaluation, judgment, opinion) or with the aid of pencil and paper but for the recitation of generic computer components. For example, a human can determine the masking type by ranking the masking types based on accuracy and selecting the top ranked (or best) masking type.
Accordingly, at Step 2A, Prong One, the claim is directed to an abstract idea.
Step 2A, Prong Two
The claim does not include additional elements that integrate the judicial exception into a practical application.
Step 2B
The claim does not include additional elements that are sufficient to amount to “significantly more” to the judicial exception.
As to claim 10,
Step 2A, Prong One
The claim recites in part:
executing the inference processing on a basis of output data output by the model by using data in which the masking type determined on a basis of a predetermined criterion is masked as an input to the block of the model.
As drafted and under its broadest reasonable interpretation, these limitation covers performance of the limitation in the mind (including an observation, evaluation, judgment, opinion) or with the aid of pencil and paper but for the recitation of generic computer components. For example, a human can determine combinations of new input data based on the rules and patterns of past input data.
Accordingly, at Step 2A, Prong One, the claim is directed to an abstract idea.
Step 2A, Prong Two
The claim does not include additional elements that integrate the judicial exception into a practical application.
Step 2B
The claim does not include additional elements that are sufficient to amount to “significantly more” to the judicial exception.
As to claim 11,
Step 2A, Prong One
The claim recites in part:
executing the inference processing on a basis of output data output by the model by using data in which the masking type determined according to a purpose of the inference processing is masked as an input to the block of the model.
As drafted and under its broadest reasonable interpretation, these limitation covers performance of the limitation in the mind (including an observation, evaluation, judgment, opinion) or with the aid of pencil and paper but for the recitation of generic computer components. For example, a human can determine combinations of new input data based on the rules and patterns of past input data.
Accordingly, at Step 2A, Prong One, the claim is directed to an abstract idea.
Step 2A, Prong Two
The claim does not include additional elements that integrate the judicial exception into a practical application.
Step 2B
The claim does not include additional elements that are sufficient to amount to “significantly more” to the judicial exception.
As to claim 12,
Step 2A, Prong One
The claim recites in part:
executing the inference processing on a basis of output data output by the model by using data in which the masking type determined according to a user who is a target of the inference processing is masked as an input to the block of the model.
As drafted and under its broadest reasonable interpretation, these limitation covers performance of the limitation in the mind (including an observation, evaluation, judgment, opinion) or with the aid of pencil and paper but for the recitation of generic computer components. For example, a human can determine combinations of new input data based on the rules and patterns of past input data.
Accordingly, at Step 2A, Prong One, the claim is directed to an abstract idea.
Step 2A, Prong Two
The claim does not include additional elements that integrate the judicial exception into a practical application.
Step 2B
The claim does not include additional elements that are sufficient to amount to “significantly more” to the judicial exception.
Claim 13 has similar limitations as claim 1. Therefore, the claim is rejected for the same reasons as above.
Claim 14 has similar limitations as claim 1. Therefore, the claim is rejected for the same reasons as above.
As to claim 18,
Step 2A, Prong One
The claim recites in part:
wherein the masking type is determined for the block using a masking list that specifies which type is to be masked among the types used as inputs of the block.
As drafted and under its broadest reasonable interpretation, these limitation covers performance of the limitation in the mind (including an observation, evaluation, judgment, opinion) or with the aid of pencil and paper but for the recitation of generic computer components. For example, a human can determine the masking type by ranking the masking types based on accuracy and selecting the top ranked (or best) masking type.
Accordingly, at Step 2A, Prong One, the claim is directed to an abstract idea.
Step 2A, Prong Two
The claim does not include additional elements that integrate the judicial exception into a practical application.
Step 2B
The claim does not include additional elements that are sufficient to amount to “significantly more” to the judicial exception.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claim(s) 1 - 8, 13, 14, and 18 is/are rejected under 35 U.S.C. 103 as being unpatentable over Gotou (US 2019/0034803) in view of Howard (US 2020/0057964) and in further view of Sequeira et al (US 2022/0067525).
As to claim 1, Gotou figure 1 shows and teaches an information processing method executed by a computer (paragraph [0097]…the information processing apparatus 100 may be one computer or may be realized by a plurality of computers performing processing in a distributed manner), the information processing method comprising:
acquiring learning data used for learning of a model comprising a neural network having an input layer, at least one intermediate layer, and an output layer, the model having at least one block, the learning data including a plurality of types of information (paragraph [0031]…Sensor data 60 are measured values (sensor data) outputted by sensors that measure various states of an environment. For example, a temperature sensor, a displacement sensor, a pressure sensor, a current sensor, a speed sensor, an acceleration sensor, a camera, an illuminance sensor, a microphone, a smell sensor, a length measure, and the like output the sensor data 60. The present embodiment includes two or more sensors, and the sensors each output different types of sensor data. For example, the sensor data 60 can be a combination of audio data outputted by a microphone, acceleration data outputted by an acceleration sensor, image data outputted by a camera for capturing an image of a top surface of a workpiece, and image data outputted by a camera for capturing an image of a side surface of the workpiece. The information processing apparatus 100 receives the sensor data 60 through an interface 18, and passes the sensor data 60 to the CPU 11; paragraph [0056]…the input data control section 140 determines selection rules for sensor data that are inputted to the machine learning apparatus 300, and gives the determined selection rules to the input data selection section 110. The input data control section 140 may retain some typical selection rules in advance and include a user interface that a user uses to employ a desired selection rule among the typical selection rules. A selection rule means a rule for selecting m (m≤n) types of sensor data from n types of the sensor data 60 inputted to the information processing apparatus 100 and generating a plurality of different combinations. A simplest selection rule is to randomly determine combinations of sensor data. At this time, combinatorial optimization may be performed based on a result of performance evaluation described below, using a genetic algorithm ; paragraph [0047]… FIG 4B schematically shows a model of a three-layer neural network configured by combining neurons shown in FIG. 4A) (Examiner’s Note: “sensor data that are inputted to the machine learning apparatus 300” reads on “acquiring learning data used for learning of a model having at least one block”; “Sensor data 60 are measured values (sensor data) outputted by sensors that measure various states of an environment” reads on “the learning data including a plurality of types of information” ; “a model of a three-layer neural network configured by combining neurons” reads on “a model comprising a neural network having an input layer, at least one intermediate layer, and an output layer”); and
selecting a type included in data to be input to the block by processing based on a genetic algorithm in learning using the learning data, and generating the model by using data corresponding to a combination of types selected among the plurality of types as an input from the input layer to the block (paragraph [0056]…the input data control section 140 determines selection rules for sensor data that are inputted to the machine learning apparatus 300, and gives the determined selection rules to the input data selection section 110. The input data control section 140 may retain some typical selection rules in advance and include a user interface that a user uses to employ a desired selection rule among the typical selection rules. A selection rule means a rule for selecting m (m≤n) types of sensor data from n types of the sensor data 60 inputted to the information processing apparatus 100 and generating a plurality of different combinations. A simplest selection rule is to randomly determine combinations of sensor data. At this time, combinatorial optimization may be performed based on a result of performance evaluation described below, using a genetic algorithm ; paragraph [0076]… The input data selection section 110 similarly performs learning for all of randomly- generated combinations of image capture conditions. Thus, a number of learning models which is equal to the number of randomly- generated combinations of image capture conditions are generated)(Examiner’s Note: “a selection rule means a rule for selecting m (m≤n) types of sensor data from n types of the sensor data 60 inputted to the information processing apparatus 100…based on a result of performance evaluation described below, using a genetic algorithm” reads on “selecting a type included in data to be input to the block by processing based on a genetic algorithm in learning using the learning data”; “a number of learning models are generated” reads on “generating the model by using data corresponding to a combination of types selected among the plurality of types as an input to the block”).
Gotou fails to explicitly show/teach learning data used for learning of a model having at least one block to which an output from an input layer is input and executing inference processing on a basis of output data output by the model by using data
in which a masking type that is a type to be partially masked among the combination of types is masked as an input to the block of the model.
However, Howards figure 1 teaches learning data used for learning of a model having at least one block to which an output from an input layer is input (paragraph [0039]… Input Data Layer 102 may include data-capturing points from data channels 108 associated with types of data: video, image, text, audio, etc., as well as meta world data 110 and objective data 112. The data channels layer may include several stages of data retrieval and manipulation, such as: identification of input points and types for each data channel, retrieval of data and data preprocessing, and data sampling techniques and storage) (Examiner’s Note: “Input Data Layer 102 may include data-capturing points from data channels 108 associated with types of data” reads on “learning data used for learning of a model having at least one block to which an output from an input layer is input”).
Therefore, it would have been obvious for one having ordinary skill in the art, at the time the invention was made, for Gotou’s learning data used for learning of a model having at least one block to which an output from an input layer is input, as in Howard, for the purpose of automatically generating or obtaining models for generating results relevant to input problems (Taught in paragraph [0002]).
Gotou and Howards both fails to explicitly show/teach executing inference processing on a basis of output data output by the model by using data in which a masking type that is a type to be partially masked among the combination of types is masked as an input to the block of the model
However, Sequeira et al teaches executing inference processing on a basis of output data output by the model by using data in which a masking type that is a type to be partially masked among the combination of types is masked as an input to the block of the model (paragraph [0060]…In at least one embodiment, representation of neural network 104 is a neural network model. In at least one embodiment, representation of neural network 104 is stored in a standardized neural network format (e.g., an Open Neural Network Exchange (ONNX) model, Universal Framework Format (UFF) model, Caffe model, or some other suitable format). In at least one embodiment, a training framework (e.g., PyTorch, TensorFlow, MXNet, Caffe, or some other suitable training framework) generates representation of neural network 104 using an online masking technique that masks some neurons (e.g., by setting weights and gradients of masked neurons and/or masked convolutional filters to zero during training) such that at an end of a training phase, a resulting model features a number of neurons and/or filters whose weights are all exactly zeroes).
Therefore, it would have been obvious for one having ordinary skill in the art, at the time the invention was made, for Gotou’s executing inference processing on a basis of output data output by the model by using data in which a masking type that is a type to be partially masked among the combination of types is masked as an input to the block of the model, as in Sequeira et al, for the purpose of reducing size of networks.
As to claim 2, Howard teaches determining the combination of types partially used as an input to the block at a time of inference using the model (paragraph [0040]…model selector 114 identify a set of methods and operations from model repository 116 to apply on the input data in relation to intelligence inferring and pattern determination. Such mechanisms may include the stages such as a Critic-Selector Mechanism, which may be based on combining input data types from data channels 108, meta world data 110, such as processing history, and objective data 112, including research knowledge and situational context to determine what is the most appropriate Artificial Intelligence (AI) model for existing data and how the system should manage the processing resources, be it models or computing infrastructure. Such mechanisms may further include data processing using AI/ML algorithms in pipelines and a models training loop and transfer learning mechanism)(Examiner’s Note: “Model selector 114 identify a set of methods and operations from model repository 116 to apply on the input data in relation to intelligence inferring and pattern determination. Such mechanisms may include the stages such as a Critic-Selector Mechanism, which may be based on combining input data types from data channels 108, meta world data 110, such as processing history, and objective data 112” reads on “determining the combination of types partially used as an input to the block at a time of inference using the model”).
It would have been obvious for determining the combination of types partially used as an input to the block at a time of inference using the model, for the same reasons as above.
As to claim 3, Howard teaches determining a type to be masked at a time of inference using the model among the combination of types (paragraph [0053]… Embodiments may provide closed-loop, activity-guided control of neural circuit dynamics using optical and electrical stimulation, while simultaneously factoring in observed dynamics in a principled way. This may provide a powerful strategy for causal investigation of neural circuitry. In particular, observing and feeding back the effects of circuit interventions on physiologically relevant timescales is valuable for directly testing whether inferred models of dynamics, connectivity, or causation is as accurate in vivo) (Examiner’s Note:” a powerful strategy for causal investigation of neural circuitry. In particular, observing and feeding back the effects of circuit interventions on physiologically relevant timescales is valuable for directly testing whether inferred models of dynamics, connectivity, or causation is as accurate in vivo” reads on “determining a type to be masked at a time of inference using the model among the combination of types”).
It would have been obvious for, determining a type to be masked at a time of inference using the model among the combination of types, for the same reasons as above.
As to claim 4, modified Gotou teaches generating the model in which the combination of types included in the data input from the input layer to the block is determined by combination optimization based on the genetic algorithm (paragraph [0094]…a most typical setting rule for the hyperparameter is to randomly set the hyperparameter. Combinatorial optimization may be performed using a genetic algorithm as performed on combinations of sensor data in Embodiment 1. In this case, as shown in the flowchart in FIG. 8, the input data selection section 110 sequentially executes the learning process described in Embodiment 1 for each of respective learners constructed using randomly-selected hyperparameters of the neural network. Also, the performance evaluation section 130 sequentially executes the performance evaluation process described in Embodiment 1 for each of respective recognizers constructed using the hyperparameters)(Examiner’s Note: “Combinatorial optimization may be performed using a genetic algorithm as performed on combinations of sensor data in Embodiment 1” reads on “generating the model in which the combination of types included in the data input from the input layer to the block is determined by combination optimization based on the genetic algorithm”).
As to claim 5, modified Gotou teaches generating the model in which the combination of types included in the data input from the input layer to the block is determined by a search based on the genetic algorithm (paragraph [0095]… This information processing apparatus of this Embodiment 2 can identify not only an optimal combination of sensor data but also an optimal hyperparameter of the neural network. Thus, the combination of sensor data can be optimized more. There are cases where because of the bad design of the neural network, a satisfactory recognition result cannot be obtained even though essentially needed sensor data are fully given. In such cases, if the structure of the neural network is exhaustively changed according to the present embodiment, a search space is widened, and an optimal combination of sensor data can be discovered)(Examiner’s Note: “if the structure of the neural network is exhaustively changed according to the present embodiment, a search space is widened, and an optimal combination of sensor data can be discovered” reads on “generating the model in which the combination of types included in the data input from the input layer to the block is determined by a search based on the genetic algorithm”).
As to claim 6, Sequeira et al teaches determining the masking type at a time of learning the model by measuring accuracy of the model while changing a masking type candidate that is a type among the combination of types, and determining, as the masking type, the masking type candidate for which the accuracy is best
(paragraph [0064]… bias terms are present in representation of neural network 104, and were not masked during training, even for masked neurons and/or convolutional filters. In at least one embodiment, even if a bias term is zero, either though masking of bias term during training, or through a trained zero value, optimizer 206 takes corresponding output plane into account during a graph optimization. In at least one embodiment, optimizer 206 propagates bias terms in a manner that accounts for cases where a bias term of zero does not result in a zero value in an output plane, such as where a pruned neuron is followed by an activation layer that transforms zeroes into non-zeroes (e.g., a sigmoid activation function), where a pruned neuron is followed by an offset layer (e.g., batch normalization), and/or where an output feature map of a pruned neuron is added to another feature map, element-wise (e.g., for a residual layer in a residual neural network)).
It would have been obvious for determining the masking type at a time of learning the model by measuring accuracy of the model while changing a masking type candidate that is a type among the combination of types, and determining, as the masking type, the masking type candidate for which the accuracy is best, for the same reasons as above.
As to claim 7, Howard teaches executing the inference processing on a basis of output data output by the model by using input data corresponding to the combination of types determined as an input to the block of the mode (paragraph [0040]…model selector 114 identify a set of methods and operations from model repository 116 to apply on the input data in relation to intelligence inferring and pattern determination. Such mechanisms may include the stages such as a Critic-Selector Mechanism, which may be based on combining input data types from data channels 108, meta world data 110, such as processing history, and objective data 112, including research knowledge and situational context to determine what is the most appropriate Artificial Intelligence (AI) model for existing data and how the system should manage the processing resources, be it models or computing infrastructure. Such mechanisms may further include data processing using AI/ML algorithms in pipelines and a models training loop and transfer learning mechanism)(Examiner’s Note: “Model selector 114 identify a set of methods and operations from model repository 116 to apply on the input data in relation to intelligence inferring and pattern determination. Such mechanisms may include the stages such as a Critic-Selector Mechanism, which may be based on combining input data types from data channels 108, meta world data 110, such as processing history, and objective data 112” reads on “executing the inference processing on a basis of output data output by the model by using input data corresponding to the combination of types determined as an input to the block of the model”).
It would have been obvious for determining the combination of types partially used as an input to the block at a time of inference using the model, for the same reasons as above.
As to claim 8, Howard teaches executing the inference processing on a basis of output data output by the model by using input data corresponding to only a part of the combination of types determined as an input to the block of the mode (paragraph [0040]…model selector 114 identify a set of methods and operations from model repository 116 to apply on the input data in relation to intelligence inferring and pattern determination. Such mechanisms may include the stages such as a Critic-Selector Mechanism, which may be based on combining input data types from data channels 108, meta world data 110, such as processing history, and objective data 112, including research knowledge and situational context to determine what is the most appropriate Artificial Intelligence (AI) model for existing data and how the system should manage the processing resources, be it models or computing infrastructure. Such mechanisms may further include data processing using AI/ML algorithms in pipelines and a models training loop and transfer learning mechanism)(Examiner’s Note: “Model selector 114 identify a set of methods and operations from model repository 116 to apply on the input data in relation to intelligence inferring and pattern determination. Such mechanisms may include the stages such as a Critic-Selector Mechanism, which may be based on combining input data types from data channels 108, meta world data 110, such as processing history, and objective data 112” reads on “executing the inference processing on a basis of output data output by the model by using input data corresponding to only a part of the combination of types determined as an input to the block of the mode”).
It would have been obvious for determining the combination of types partially used as an input to the block at a time of inference using the model, for the same reasons as above.
As to claim 9, Sequeira et al teaches determining the masking type by a search based on the genetic algorithm performed in consideration of a type to be masked (paragraph [0060]…In at least one embodiment, representation of neural network 104 is a neural network model. In at least one embodiment, representation of neural network 104 is stored in a standardized neural network format (e.g., an Open Neural Network Exchange (ONNX) model, Universal Framework Format (UFF) model, Caffe model, or some other suitable format). In at least one embodiment, a training framework (e.g., PyTorch, TensorFlow, MXNet, Caffe, or some other suitable training framework) generates representation of neural network 104 using an online masking technique that masks some neurons (e.g., by setting weights and gradients of masked neurons and/or masked convolutional filters to zero during training) such that at an end of a training phase, a resulting model features a number of neurons and/or filters whose weights are all exactly zeroes).
It would have been obvious for determining the masking type by a search based on the genetic algorithm performed in consideration of a type to be masked, for the same reasons as above.
As to claim 10, Sequeira et al teaches executing the inference processing on a basis of output data output by the model by using data in which the masking type determined on a basis of a predetermined criterion is masked as an input to the block of the model (paragraph [0060]…In at least one embodiment, representation of neural network 104 is a neural network model. In at least one embodiment, representation of neural network 104 is stored in a standardized neural network format (e.g., an Open Neural Network Exchange (ONNX) model, Universal Framework Format (UFF) model, Caffe model, or some other suitable format). In at least one embodiment, a training framework (e.g., PyTorch, TensorFlow, MXNet, Caffe, or some other suitable training framework) generates representation of neural network 104 using an online masking technique that masks some neurons (e.g., by setting weights and gradients of masked neurons and/or masked convolutional filters to zero during training) such that at an end of a training phase, a resulting model features a number of neurons and/or filters whose weights are all exactly zeroes).
It would have been obvious for executing the inference processing on a basis of output data output by the model by using data in which the masking type determined on a basis of a predetermined criterion is masked as an input to the block of the model, for the same reasons as above.
As to claim 11, Sequeira et al teaches executing the inference processing on a basis of output data output by the model by using data in which the masking type determined according to a purpose of the inference processing is masked as an input to the block of the model (paragraph [0060]…In at least one embodiment, representation of neural network 104 is a neural network model. In at least one embodiment, representation of neural network 104 is stored in a standardized neural network format (e.g., an Open Neural Network Exchange (ONNX) model, Universal Framework Format (UFF) model, Caffe model, or some other suitable format). In at least one embodiment, a training framework (e.g., PyTorch, TensorFlow, MXNet, Caffe, or some other suitable training framework) generates representation of neural network 104 using an online masking technique that masks some neurons (e.g., by setting weights and gradients of masked neurons and/or masked convolutional filters to zero during training) such that at an end of a training phase, a resulting model features a number of neurons and/or filters whose weights are all exactly zeroes).
It would have been obvious for executing the inference processing on a basis of output data output by the model by using data in which the masking type determined according to a purpose of the inference processing is masked as an input to the block of the model, for the same reasons as above.
As to claim 12, Sequeira et al teaches executing the inference processing on a basis of output data output by the model by using data in which the masking type determined according to a user who is a target of the inference processing is masked as an input to the block of the model (paragraph [0060]…In at least one embodiment, representation of neural network 104 is a neural network model. In at least one embodiment, representation of neural network 104 is stored in a standardized neural network format (e.g., an Open Neural Network Exchange (ONNX) model, Universal Framework Format (UFF) model, Caffe model, or some other suitable format). In at least one embodiment, a training framework (e.g., PyTorch, TensorFlow, MXNet, Caffe, or some other suitable training framework) generates representation of neural network 104 using an online masking technique that masks some neurons (e.g., by setting weights and gradients of masked neurons and/or masked convolutional filters to zero during training) such that at an end of a training phase, a resulting model features a number of neurons and/or filters whose weights are all exactly zeroes).
It would have been obvious for executing the inference processing on a basis of output data output by the model by using data in which the masking type determined according to a user who is a target of the inference processing is masked as an input to the block of the model, for the same reasons as above.
Claim 13 has similar limitations as claim 1. Therefore, the claim is rejected for the same reasons as above.
Claim 14 has similar limitations as claim 1. Therefore, the claim is rejected for the same reasons as above.
As to claim 18, Sequeira et al teaches wherein the masking type is determined for the block using a masking list that specifies which type is to be masked among the types used as inputs of the block (paragraph [0060]…In at least one embodiment, representation of neural network 104 is a neural network model. In at least one embodiment, representation of neural network 104 is stored in a standardized neural network format (e.g., an Open Neural Network Exchange (ONNX) model, Universal Framework Format (UFF) model, Caffe model, or some other suitable format). In at least one embodiment, a training framework (e.g., PyTorch, TensorFlow, MXNet, Caffe, or some other suitable training framework) generates representation of neural network 104 using an online masking technique that masks some neurons (e.g., by setting weights and gradients of masked neurons and/or masked convolutional filters to zero during training) such that at an end of a training phase, a resulting model features a number of neurons and/or filters whose weights are all exactly zeroes).
It would have been obvious for wherein the masking type is determined for the block using a masking list that specifies which type is to be masked among the types used as inputs of the block, for the same reasons as above.
Response to Arguments
Applicant's arguments filed 8/13/2026 have been fully considered but they are not persuasive.
Claim Rejections - 35 USC § 101
The 101 Rejection still has not been overcome. The claims are abstract and the steps in the claims can be completed with a mental process and/or generic computer components. Additionally, the steps in the claims do not describe an improvement of technology in any way.
The applicant argues:
Claims 1-14 stand rejected under 35 U.S.C. § 101 as allegedly being directed to an abstract idea without significantly more. The Examiner has alleged that the selecting and generating steps of claim 1 can be performed in the mind or with the aid of pencil and paper, and that the acquiring step amounts to insignificant extra-solution activity. Applicant respectfully traverses this rejection.
Applicant submits that claims 1, 13, and 14, as amended, integrate any alleged abstract idea into a practical application and recite significantly more than an abstract idea. As amended, claim 1 recites acquiring learning data used for learning of a model "comprising a neural network having an input layer, at least one intermediate layer, and an output layer," selecting a type included in data to be input to the block by processing based on a genetic algorithm, generating the model by using data corresponding to a combination of types selected among the plurality of types as an input from the input layer to the block, and "executing inference processing on a basis of output data output by the model by using data in which a masking type that is a type to be partially masked among the combination of types is masked as an input to the block of the model." Claims 13 and recite corresponding features.
These features do not recite a mental process. Generating a neural-network model - having an input layer, intermediate layer(s), and an output layer - by iterative genetic-algorithm optimization, and then executing inference processing through the neural network by masking a type among the combination of types as an input to the block, cannot practically be performed in the human mind or with pencil and paper. Moreover, the amended claims integrate any alleged abstract idea into a practical application and amount to significantly more because they recite a specific improvement to machine-learning technology: the trained neural network is operated to perform inference on partial input by masking a type among the combination of types, SO the model can flexibly use input data and produce output using only a part of the input types, reducing the input data and computational resources required at inference. See As-Filed Specification, paragraph [0090]. In addition, as described in the specification, reducing the number of dimensions of the feature space corresponding to the model reduces the approximation error and improves model accuracy. See id., paragraphs [0201]-[0202]; see also id., paragraph [0004].
Accordingly, Applicant respectfully submits that claims 1, 13, and 14, as amended, are patent-eligible under 35 U.S.C. § 101, and dependent claims 2-12 are likewise patent-eligible at least by virtue of their dependency. Applicant respectfully requests withdrawal of the rejection under 35 U.S.C. § 101.
The examiner disagrees. The examiner respectfully disagrees with the applicant’s position, as the arguments presented rely on limitations that are neither explicitly recited in the claims nor reasonably inferred from them. At no point in the pending claims does the applicant assert, describe, or even suggest the limitations of:
Generating a neural-network model - having an input layer, intermediate layer(s), and an output layer - by iterative genetic-algorithm optimization
the trained neural network is operated to perform inference on partial input by masking a type among the combination of types, SO the model can flexibly use input data and produce output using only a part of the input types, reducing the input data and computational resources required at inference
reducing the number of dimensions of the feature space corresponding to the model reduces the approximation error and improves model accuracy
Rather, the applicant appears to have introduced this language as part of the argument, but such a limitation cannot be read into the claims when it is not supported by the actual claim language. Without clear support in the claim language the examiner cannot give weight to arguments premised on these alleged limitations.
Although the neural-network operations may not be performed mentally, this along does not establish eligibility. The recited selection and masking steps have not been show to improve how the model operates, rather than merely improving data-processing results. Accordingly, the asserted benefits do not establish integration into a practical application or significantly more.
Whether the claim invokes computers or other machinery merely as a tool to perform an existing process. Use of a computer or other machinery in its ordinary capacity for information processing or other tasks (e.g., to receive, store, or transmit data) or simply adding a general purpose computer or computer components after the fact to an abstract idea does not integrate a judicial exception into a practical application or provide significantly more. See Affinity Labs v. DirecTV, 838 F.3d 1253, 1262, 120 USPQ2d 1201, 1207 (Fed. Cir. 2016) (cellular telephone); TLI Communications LLC v. AV Auto, LLC, 823 F.3d 607, 613, 118 USPQ2d 1744, 1748 (Fed. Cir. 2016) (computer server and telephone unit). Similarly, "claiming reducing errors or improved accuracy” is inherent with applying the abstract idea on a computer" does not integrate a judicial exception into a practical application or provide an inventive concept. Intellectual Ventures I LLC v. Capital One Bank (USA), 792 F.3d 1363, 1367, 115 USPQ2d 1636, 1639 (Fed. Cir. 2015). In contrast, a claim that purports to improve computer capabilities or to improve an existing technology may integrate a judicial exception into a practical application or provide significantly more. McRO, Inc. v. Bandai Namco Games Am. Inc., 837 F.3d 1299, 1314-15, 120 USPQ2d 1091, 1101-02 (Fed. Cir. 2016); Enfish, LLC v. Microsoft Corp., 822 F.3d 1327, 1335-36, 118 USPQ2d 1684, 1688-89 (Fed. Cir. 2016). See MPEP §§ 2106.04(d)(1) and 2106.05(a) for a discussion of improvements to the functioning of a computer or to another technology or technical field.
Claim Rejections - 35 USC § 103
Gotou and Howards both fail to explicitly show/teach executing inference processing on a basis of output data output by the model by using data in which a masking type that is a type to be partially masked among the combination of types is masked as an input to the block of the model
However, Sequeira et al teaches executing inference processing on a basis of output data output by the model by using data in which a masking type that is a type to be partially masked among the combination of types is masked as an input to the block of the model (paragraph [0060]…In at least one embodiment, representation of neural network 104 is a neural network model. In at least one embodiment, representation of neural network 104 is stored in a standardized neural network format (e.g., an Open Neural Network Exchange (ONNX) model, Universal Framework Format (UFF) model, Caffe model, or some other suitable format). In at least one embodiment, a training framework (e.g., PyTorch, TensorFlow, MXNet, Caffe, or some other suitable training framework) generates representation of neural network 104 using an online masking technique that masks some neurons (e.g., by setting weights and gradients of masked neurons and/or masked convolutional filters to zero during training) such that at an end of a training phase, a resulting model features a number of neurons and/or filters whose weights are all exactly zeroes).
Therefore, Gotou in view of Howards and in further view of Sequeira et al teaches all the limitations as claimed.
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 BRANDON S COLE whose telephone number is (571)270-5075. The examiner can normally be reached Mon - Fri 7:30pm - 5pm EST (Alternate Friday's Off).
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, Omar Fernandez can be reached at 571-272-2589. 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.
/BRANDON S COLE/ Primary Examiner, Art Unit 2128
