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 .
This action is made final.
Claims 1-21 are pending. Claims 1, 17 and 18 are independent claims.
Response to Arguments
Applicant’s arguments, dated 1/20/2026, regarding the 35 U.S.C. 101 rejections of the previous office action have been fully considered but are unpersuasive. Applicant argues that the use of the attention mechanism enables improved expressiveness of environment features. The examiner argues that the generation and application of the attention weights are a mental process and mathematical calculation, respectively, and cannot alone provide the technological improvement. See MPEP2106.05(a), ¶6, “the judicial exception alone cannot provide the improvement”. The examiner argues that the claimed additional elements, when considered alone or in combination, fail to provide a technological improvement because they amount to inputting/outputting from a generally recited network.
Applicant’s arguments, dated 1/20/2026, regarding the 35 U.S.C. 103 rejections of the previous office action have been fully considered. Due to the amendments, the scope of the claims has changed and new grounds of rejection are applied – see the updated rejection below.
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 1-21 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Regarding claim 1:
Step 1: This part of the eligibility analysis evaluates whether the claim falls within any statutory category. See MPEP 2106.03. Claim 1 recites: A computer-implemented method for controlling an agent interacting with an environment to perform a task, the method comprising… Claim 1 is directed to a method (Step 1: YES).
Step 2A prong 1: Does the claim recite a judicial exception? Claim 1 recites: processing the observation… to receive as input the observation and to generate as output an… representation of the observation that comprises a respective feature vector for each of a plurality of spatially distinct portions of the observation, wherein each respective feature vector has a plurality of dimensions (generating multiple feature vectors for spatially distinct portions of an observation is a mental process, i.e., dividing an image into quadrants and identifying details like color and shape information within each quadrant); for each of a plurality of subschema recurrent neural networks: generating a respective attention weight for each of the plurality of dimensions in the respective feature vector for each of the plurality of spatially distinct portions of the observation (generating an attention weight for a feature vector is a mental process, i.e., evaluating the importance of identified features)… generating an attended encoder representation, comprising applying, to the respective feature vector for each of the plurality of spatially distinct portions of the observation, the respective attention weights generated for each of the plurality of dimensions in the respective feature vector (applying attention weights to their respective feature vectors involves mathematical calculations like an element-wise product). These steps can be performed mentally or are mathematical calculations (Step 2A prong 1: YES).
Step 2A prong 2: Does the claim recite additional elements? Do those additional elements, considered individually and in combination, integrate the judicial exception into a practical application? Claim 1 recites: receiving an observation that characterizes a current state of the environment… using an encoder neural network configured… encoder… from at least a subschema hidden state of the subschema recurrent neural network, and updating the subschema hidden state using at least the attended encoder representation; and selecting an action to be performed by the agent in response to the observation using the updated subschema hidden states of the plurality of subschema recurrent neural networks. Receiving an observation that characterizes a current state of the environment and selecting an action to be performed as a result of the method are extra-solution activity of data gathering/outputting respectively that does not add a meaningful limitation to the agent training method. Using an encoder neural network to output an encoder representation, obtaining attention weights from at least a subschema hidden state, and updating subschema hidden states with the output of said encoder are attempts to use the neural network models by merely applying the abstract idea (i.e., perform the math or mental processes) without placing any limits on how the neural network model operates. Further, the claim omits any details as to how the neural network model solves a technical problem and instead recites only the idea of a solution or outcome (see MPEP 2106.05(f)). Thus, the limitation represents no more than mere instructions to implement the abstract idea which is equivalent to adding the words “apply it” to the recited judicial exception (Step 2A prong 2: NO).
Step 2B: These elements are recited at such a high level of generality that they fail to integrate the abstract idea into a practical application, since they only amount to data gathering or outputting without significantly more (MPEP 2106.05(g)) or provide nothing more than mere instructions to implement an abstract idea on a generic computer (MPEP 2106.05(f)). These limitations, taken either alone or in combination, fail to provide an inventive concept (Step 2B: NO). Thus, the claim is not patent eligible.
Regarding claims 2-16 and 21, they recite limitations which further narrow the abstract idea by specifying more details of the mental and mathematical process that occurs (Claim 2, inputting an image as the observation and analyzing different portions of the image is limiting the field of use to image analysis without significantly more; Claim 3, inputting an audio and analyzing different frequency bands is limiting the field of use to audio analysis without significantly more; Claim 4, inputting proprioception information and analyzing different robot body parts is limiting the field of use to robot control without significantly more; Claim 5, determining a subschema query is a mental process, i.e., reshaping and combining vectors; Claim 6, determining a subschema query from task description text is recited at a high level of generality and provides nothing more than mere instructions to implement an abstract idea on a generic computer; Claim 7, applying feature coefficient weights to the subschema query is a mathematical calculation, i.e., element-wise multiplication; Claim 8, applying attention weights via an element-wise product is a mathematical calculation; Claim 9, applying an attention mechanism to hidden states of neural networks is recited at a high level of generality and provides nothing more than mere instructions to implement an abstract idea on a generic computer, while obtaining information from other neural networks is insignificant extra-solution activity of data gathering; Claim 10, applying the attention mechanism to a null vector is recited at a high level of generality and provides nothing more than mere instructions to implement an abstract idea on a generic computer; Claim 11, updating the subschema hidden state is recited at a high level of generality and provides nothing more than mere instructions to implement an abstract idea on a generic computer; Claim 12, processing a policy input to generate an action selection to be performed by the agent is recited at a high level of generality and provides nothing more than mere instructions to implement an abstract idea on a generic computer; Claim 13, training the action selection policy neural network is recited at a high level of generality and provides nothing more than mere instructions to implement an abstract idea on a generic computer; Claim 14, determining trained parameter values for the encoder neural network and subschema RNNs is recited at a high level of generality and provides nothing more than mere instructions to implement an abstract idea on a generic computer; Claim 15, using the method to perform an object manipulation task or environment navigation task is limiting the field of use without significantly more; Claim 16, operating a mechanical agent in a real-world environment using data gathered by sensors is limiting the field of use without significantly more; Claim 21, including color, shape, pattern or other information as dimensions in a feature vector is a mental process – i.e., a human can identify and record color information for parts of an image).
Regarding claim 17, it is an apparatus that recites limitations similar to claim 1 and is rejected on the same grounds – see above.
Regarding claim 18, it is a system (i.e., an apparatus) that recites limitations similar to claim 1 and is rejected on the same grounds – see above.
Regarding claims 19 and 20, they teach limitations similar to claims 5 and 6 respectively and are rejected on the same grounds – see above.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 1, 2, 5, 7-9, 11 and 17-19 is/are rejected under 35 U.S.C. 103 as being unpatentable over Mott et al. (“Towards Interpretable Reinforcement Learning Using Attention Augmented Agents”, 2019, INCLUDED IN IDS), herein Mott, in view of Liu et al. (WO 2022031232 A1), herein Liu, and Yu et al. (US 11551042 B1), herein Yu.
Regarding claim 1, Mott teaches: A computer-implemented method for controlling an agent interacting with an environment to perform a task (pg. 8, Conclusion, We have applied an attention mechanism to an agent trained with reinforcement learning on the ATARI environment), the method comprising: receiving an observation that characterizes a current state of the environment; processing the observation using an encoder neural network configured to receive as input the observation and to generate as output an encoder representation of the observation that comprises a respective feature vector for each of a plurality of spatially distinct portions of the observation, wherein each respective feature vector has a plurality of dimensions (pg. 2, Model, An observation… is passed through a “vision core”… such as a ConvLSTM [4], which produces an output tensor… the vision core output is then split along the channel dimension into two tensors: the “Keys” tensor… and the “Values” tensor V… we concatenate a spatial basis — a fixed tensor… which encodes spatial locations); for… a… recurrent neural network: generating a respective attention weight… from at least a… hidden state of the… recurrent neural network (pg. 2, Fig. 1, A recurrent network at the top sends its state from the previous time-step… produce an attention map for the query), generating an attended encoder representation, comprising applying, to the respective feature vector for each of the plurality of spatially distinct portions of the observation, the respective attention weights (pg. 2, Fig. 1, The attention map is broadcast along the channel dimension, point-wise multiplied with the values tensor and the result is then summed across space to produce an answer vector. This answer is sent to the top LSTM as input to produce the output and next state of the LSTM)… and updating the… hidden state using at least the attended encoder representation (pg. 2, Fig. 1, The attention map is broadcast along the channel dimension, point-wise multiplied with the values tensor and the result is then summed across space to produce an answer vector. This answer is sent to the top LSTM as input to produce the output and next state of the LSTM); and selecting an action to be performed by the agent in response to the observation using the updated… hidden states of the… recurrent neural network (pg. 1, Abstract, The output of the attention mechanism allows direct observation of the information used by the agent to select its actions).
Mott fails to teach: attention weights for each of the plurality of dimensions in the respective feature vector for each of the plurality of spatially distinct portions of the observation… generated for each of the plurality of dimensions in the respective feature vector…
However, in the same field of endeavor, Liu teaches: attention weights for each of the plurality of dimensions in the respective feature vector for each of the plurality of spatially distinct portions of the observation… generated for each of the plurality of dimensions in the respective feature vector… (¶65, This is reflected by the B-Conv layer 200 having a first MLP 201 for deriving attention weight information from difference in 3D position (of a neighbourhood point to a respective query point), a second MLP 202 for deriving attention weight information from difference in colour, a third MLP 203 for deriving attention weight information from difference in geometric features and a fourth MLP 204 for deriving attention weight information from difference in propagated features (input from a preceding B-Conv layer). These MLPs 201 to 204 can be seen as partial attention value determining perceptrons since they generate “partial” attention values which are used to generate the final attention values of attentional weight 208).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to use respective attention weights for different dimensions of a feature vector as disclosed by Liu in the method disclosed by Mott to discover additional insight in input data (¶70, The first stage uses the four MLPs 201, 202, 203, 204 and extracts high level features unique to the four (difference) spaces respectively).
Mott in view of Liu fails to teach: each of a plurality of subschema recurrent neural networks… subschema hidden state, subschema hidden states of the plurality of subschema recurrent neural networks.
However, in the same field of endeavor, Yu teaches: each of a plurality of subschema recurrent neural networks… subschema hidden state, subschema hidden states of the plurality of subschema recurrent neural networks (Fig. 8, plurality of RNNs that focuses on a different part of the input).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to use multiple subschema recurrent neural networks as disclosed by Yu in the method disclosed by Mott in view of Liu to accurately analyze content with multiple distinct parts (Col. 2, line 36, to efficiently and more accurately generate entity-level sentiment classifications for multimodal messages… the left and right contexts are further processed using an attention mechanism to capture the most important context information).
Regarding claim 2, Mott further teaches: The method of claim 1, wherein the observation comprises an image, and wherein the plurality of spatially distinct portions of the observation correspond to different spatial positions of the image (pg. 2, Model, An observation X… at time t (here an RGB frame of height H and width W) is passed through a “vision core”. The vision core is a multi-layer convolutional network… followed by a recurrent layer… which produces an output tensor).
Regarding claim 5, Mott further teaches: The method of claim 1, further comprising, for each of the plurality of subschema recurrent neural networks: determining a subschema query from (i) the subschema hidden state of the subschema recurrent neural network and one or more of: (ii) a preceding action performed by the agent in response to a preceding observation characterizing a preceding state of the environment state that precedes the current state of the environment state, or (iii) a preceding reward received in response to the agent performing the preceding action (pg. 4, Model, ¶ 4, The LSTM sends its state… previous time step t − 1 into a “Query Network”. The query network… with parameters ψ whose output is reshaped into N query vectors – and – pg. 4, Analysis and Results, Another (fully connected) LSTM… takes as input the query and answer vectors, the previous reward and a one-hot encoding of the previous action – this input comprises at least a query vector derived from a previous LSTM state, the previous reward, and the previous action).
Regarding claim 7, Mott further teaches: The method of claim 5, wherein generating the respective attention weight for each of the plurality of dimensions comprises: generating the respective attention weight for each of the plurality of dimensions based on applying one or more sets of learnt feature coefficient weights to the subschema query (pg. 3, Model, Similar to [5], we take the inner product between each query vector qn and all spatial locations in the keys tensor K to form the n-th attention logits map – also see Eq. 4).
Regarding claim 8, Mott further teaches: The method of claim 1, wherein applying, to the respective feature vector for each of the plurality of spatially distinct portions of the observation, the respective attention weights comprises: computing an element-wise product between the respective attention weights and the respective feature vector for each of the plurality of spatially distinct portions of the observation (pg. 3, Model, Each attention map An is broadcast along the channel dimension of the values tensor V… point-wise multiplied with it and then summed across space to produce the n-th answer vector).
Regarding claim 9, Mott further teaches: The method of claim 5, further comprising, for each of the plurality of subschema recurrent neural networks: obtaining… information from the… recurrent neural networks, comprising applying an attention mechanism over the… hidden states of the… recurrent neural networks using one or more queries derived from the subschema query of the subschema recurrent neural network (pg. 3, Model, Similar to [5], we take the inner product between each query vector qn and all spatial locations in the keys tensor K to form the n-th attention logits map, also see Eq. 4 – i.e., applying an attention mechanism over the hidden state in a single RNN).
Mott in view of Liu fails to teach: shared subschema information from the subschema hidden states of other subschema recurrent neural networks in the plurality of subschema recurrent neural networks… applying an attention mechanism over the subschema hidden states… plurality of subschema recurrent neural networks.
However, in the same field of endeavor, Yu further teaches: shared subschema information from the subschema hidden states of other subschema recurrent neural networks in the plurality of subschema recurrent neural networks… applying an attention mechanism over the subschema hidden states… plurality of subschema recurrent neural networks (Fig. 8, plurality of RNNs that focuses on a different part of the input, including attention layers 807A and 807B which are applied to hidden states from the subschema RNNs. The attention layer outputs for each subnetwork are combined at the end).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to share subschema information between RNNs and apply an attention mechanism as disclosed by Yu in the method disclosed by Mott in view of Liu to accurately analyze content with multiple distinct parts (Col. 2, line 36, to efficiently and more accurately generate entity-level sentiment classifications for multimodal messages… the left and right contexts are further processed using an attention mechanism to capture the most important context information).
Regarding claim 11, Mott further teaches: The method of claim 9, wherein updating the subschema hidden state comprises updating the subschema hidden state using the attended encoder representation and the shared subschema information (pg. 2, Fig. 1 caption, The attention map is broadcast along the channel dimension, point-wise multiplied with the values tensor and the result is then summed across space to produce an answer vector. This answer is sent to the top LSTM as input to produce the output and next state of the LSTM).
Regarding clam 17, it recites similar limitations to claim 1 and is rejected on the same grounds – see above.
Regarding claim 18, it recites similar limitations to claim 1 and is rejected on the same grounds – see above.
Regarding claim 19, it recites similar limitations to claim 5 and is rejected on the same grounds – see above.
Claim(s) 3 is/are rejected under 35 U.S.C. 103 as being unpatentable over Mott in view of Liu and Yu as applied to claim 1 above, and further in view of Sun et al. (US 10460722 B1), herein Sun.
Regarding claim 3, Mott in view of Liu and Yu fails to teach: The method of claim 1, wherein the observation comprises an audio, and wherein the plurality of spatially distinct portions of the observation correspond to different frequency bands of the audio.
However, in the same field of endeavor, Sun teaches: wherein the observation comprises an audio, and wherein the plurality of spatially distinct portions of the observation correspond to different frequency bands of the audio (Col. 3, lines 53-56, a feature extractor 140 receives the digitized audio signal and produces one feature vector for each 10 milliseconds of the audio signal. In this embodiment, the element of each feature vector represents the logarithm of the energy in an audio frequency band).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to use an audio input and analyze different frequency bands as disclosed by Sun in the method disclosed by Mott in view of Liu and Yu to provide audio analysis with reduced computational cost (Col. 2, line 53, The system 100 described below is an example of a system that provides an improved trigger detection error rate within a limited computation capacity).
Claim(s) 4 is/are rejected under 35 U.S.C. 103 as being unpatentable over Mott in view of Liu and Yu as applied to claim 1 above, and further in view of Pascanu et al. (US 20190232489 A1), herein Pascanu.
Regarding claim 4, Mott in view of Liu and Yu fails to teach: The method of claim 1, wherein the observation comprises proprioception information of a robot, and wherein the plurality of spatially distinct portions of the observation correspond to different body parts of the robot.
However, in the same field of endeavor, Pascanu teaches: wherein the observation comprises proprioception information of a robot, and wherein the plurality of spatially distinct portions of the observation correspond to different body parts of the robot (¶64, The additional input can be proprioceptive data including proprioceptive features of the robotic agent. The proprioceptive features may include joint angles and velocities for each of the joints and actuators of the robotic agent – and – ¶25, In some cases, the observations characterize states of the environment using low-dimensional feature vectors that characterize the state of the environment).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to use robot proprioception input and analyze data from different parts of the robot as disclosed by Pascanu in the method disclosed by Mott in view of Liu and Yu to provide improved robot training (¶8, an action selection policy for a robotic agent can be effectively determined more quickly and using fewer computing resources than existing approaches).
Claim(s) 6, 12-16, and 20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Mott in view of Liu and Yu as applied to claims 5 and 1 above, and further in view of Hermann et al. (US 20210110115 A1), herein Hermann.
Regarding claim 6, Mott in view of Liu and Yu fails to teach: The method of claim 5, further comprising determining the subschema query from task description text that specifies the task being performed by the agent.
However, in the same field of endeavor, Hermann teaches: further comprising determining the subschema query from task description text that specifies the task being performed by the agent (¶7, The subsystem is configured to receive a current text string in the particular natural language that expresses information about a current task currently being performed by the agent. The subsystem provides the current text string as input to the language encoder model to obtain a current text embedding of the current text string).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to determine the query at least partly from text as disclosed by Hermann in the method disclosed by Mott in view of Liu and Yu to reduce the memory requirement (¶59, the system as described in this specification can reduce the use of computational resources (e.g., memory) relative to some conventional systems).
Regarding claim 12, Mott in view of Liu and Yu fails to teach: The method of claim 1, wherein selecting the action to be performed by the agent comprises: processing a policy input comprising the updated subschema hidden states of the plurality of subschema recurrent neural networks using an action selection policy neural network to generate an action selection policy output that specifies the action to be performed by the agent.
However, in the same field of endeavor, Hermann teaches: wherein selecting the action to be performed by the agent comprises: processing a policy input comprising the updated subschema hidden states of the plurality of subschema recurrent neural networks using an action selection policy neural network to generate an action selection policy output that specifies the action to be performed by the agent (¶38, The policy defining neural network module is coupled to receive data from the environment neural network module and from the task neural network module and to output action data in accordance with the policy. The action data represents an action to perform in the environment – and – ¶3, select the action to be performed by the agent in response to receiving a given observation in accordance with an output of a neural network).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to process a policy input to generate an action to be performed by an agent as disclosed by Hermann in the method disclosed by Mott in view of Liu and Yu to successfully accomplish a task (¶21, to optimize the task-specific objective).
Regarding claim 13, Mott in view of Liu and Yu fails to teach: The method of claim 12, further comprising training the action selection policy neural network through reinforcement learning to determine trained parameter values of the action selection policy neural network.
However, in the same field of endeavor, Hermann teaches: further comprising training the action selection policy neural network through reinforcement learning to determine trained parameter values of the action selection policy neural network (¶39, The reinforcement learning training module trains… the policy defining neural network module in response to reward data representing successful performance of the one or more tasks).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to train the policy network as disclosed by Hermann in the method disclosed by Mott in view of Liu and Yu to successfully accomplish a task (¶21, to optimize the task-specific objective).
Regarding claim 14, Mott in view of Liu and Yu fails to teach: The method of claim 13, further comprising determining respective trained parameter values of the encoder neural network and the plurality of subschema recurrent neural networks through reinforcement learning.
However, in the same field of endeavor, Hermann teaches: further comprising determining respective trained parameter values of the encoder neural network and the plurality of subschema recurrent neural networks through reinforcement learning (¶39, The reinforcement learning training module trains the environment neural network module, the task neural network module, and the policy defining neural network module in response to reward data representing successful performance of the one or more tasks).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to determine trained parameter values in the networks with reinforcement learning as disclosed by Hermann in the method disclosed by Mott in view of Liu and Yu to successfully accomplish a task (¶21, to optimize the task-specific objective).
Regarding claim 15, Mott in view of Liu and Yu fails to teach: The method of claim 1, wherein the task comprises one of an object manipulation task or an environment navigation task.
However, in the same field of endeavor, Hermann teaches: wherein the task comprises one of an object manipulation task or an environment navigation task (¶72, In some implementations, the environment 106 is a real-world environment and the agent 104 is a mechanical agent interacting with the real-world environment. For example, the agent 104 may be a robot interacting with the environment 106 to accomplish a specific task. As another example, the agent 104 may be an autonomous or semi-autonomous vehicle navigating through the environment).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to model an environment navigation task as disclosed by Hermann in the method disclosed by Mott in view of Liu and Yu to successfully accomplish said navigation task (¶21, to optimize the task-specific objective).
Regarding claim 16, Mott in view of Liu and Yu fails to teach: The method of claim 1, wherein the agent is a mechanical agent, the environment is a real-world environment, and the observation comprises data from one or more sensors configured to sense the real-world environment.
However, in the same field of endeavor, Hermann teaches: wherein the agent is a mechanical agent, the environment is a real-world environment, and the observation comprises data from one or more sensors configured to sense the real-world environment (¶72, In some implementations, the environment 106 is a real-world environment and the agent 104 is a mechanical agent interacting with the real-world environment… the observations 110 may be generated by or derived from sensors of the agent).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to model a real-world environment and utilize sensor data as input as disclosed by Hermann in the method disclosed by Mott in view of Liu and Yu to successfully accomplish a related task (¶21, to optimize the task-specific objective).
Regarding claim 20, it recites similar limitations to claim 6 and is rejected on the same grounds – see above.
Claim(s) 10 is/are rejected under 35 U.S.C. 103 as being unpatentable over Mott in view of Liu and Yu as applied to claim 9 above, and further in view of Xia et al. (US 20200074275 A1), herein Xia.
Regarding claim 10, Mott in view of Liu and Yu fails to teach: The method of claim 9, wherein obtaining the shared subschema information further comprises applying the attention mechanism over a null vector in addition to the subschema hidden states of the plurality of subschema recurrent neural networks.
However, in the same field of endeavor, Xia teaches: wherein obtaining the shared subschema information further comprises applying the attention mechanism over a null vector in addition to the subschema hidden states of the plurality of subschema recurrent neural networks (¶40, In some embodiments, the attention layer 702 in the first chain can use a zero-vector as its hidden state vector).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to use a null vector as disclosed by Xia in the method disclosed by Mott in view of Liu and Yu to initialize the learning process (¶40, since there is no previous hidden state).
Claim(s) 21 is/are rejected under 35 U.S.C. 103 as being unpatentable over Mott in view of Liu and Yu as applied to claim 1 above, and further in view of Shen et al. (US 9996768 B2), herein Shen.
Regarding claim 21, Mott in view of Liu and Yu fails to teach: The method of claim 1, wherein the plurality of dimensions comprises two or more of: a color dimension, a shape dimension, a spatiotemporal dimension, a structure feature dimension, or a pattern dimension.
However, in the same field of endeavor, Shen teaches: wherein the plurality of dimensions comprises two or more of: a color dimension, a shape dimension, a spatiotemporal dimension, a structure feature dimension, or a pattern dimension (Col. 4, line 63, The neural network module 110, for instance, may be configured to calculate activations of image characteristics for each of the patches. As previously described, the image characteristics may describe a variety of different characteristics, such as noise, darkness, contrast, structure, whether an alignment is upright – i.e., a color feature, structure feature, and shape feature, and – Col 5, line 4, These activations 206 may be expressed in a variety of ways. For example, for each of the patches 204, a vector having a plurality of dimensions may be generated in which each of the dimensions has a corresponding image characteristic. Thus, the vector may express an amount of each of the respective characteristics through use of the activations using the dimensions of the vector).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to use feature dimensions like color, shape or structure as disclosed by Shen in the method disclosed by Mott in view of Liu and Yu to improve model performance (Col. 2, line 61, improve consistency of training and testing).
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 HARRISON CHAN YOUNG KIM whose telephone number is (571)272-0713. The examiner can normally be reached Monday - Friday 9:00 am - 5:00 pm.
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, Cesar Paula can be reached at (571) 272-4128. 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.
/HARRISON C KIM/ Examiner, Art Unit 2145
/CESAR B PAULA/ Supervisory Patent Examiner, Art Unit 2145
Prosecution Insights