DETAILED ACTION
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on 19 February 2025 has been entered.
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 .
Response to Amendment
This action is in response to the submission filed 19 February 2025 for application 17/260,258. Claims 1, 10, and 19 have been amended. Claims 21 and 22 have been newly added. Currently claims 1, 3-10, 12-17, 19, 21, and 22 are pending and have been examined.
The objection to claim 19 has been withdrawn in view of the amendments made.
Response to Arguments
Regarding Applicant’s arguments, filed 21 January 2025, see page 6, with respect to the claim Rejections under 35 USC § 103, Applicant argues that Applicant's specification teaches adaptation of an untrained ("generic") network to a specific task using a second neural network. Specification at [0041]. This concept is neither taught nor suggested by the combinations utilized by the office.
Examiner’s Response: Applicant’s arguments have been fully considered but are not persuasive because the cited references do teach “the first neural network being adapted to a different task by the second neural network” as recited in independent claim 1 (and similarly in independent claim 10). The abstract of Reference Thaler clearly states that The device may also include a reciprocal feed back connection from the output of the second neural network to the first neural network to further influence and change what takes place in the aforesaid neural network. The phrase “being adapted” is a broad term and under broadest reasonable interpretation, Examiner notes that “change what takes place in the aforesaid neural network” corresponds to the first neural network being adapted to a different task by the second neural network.
Hence, the cited references teach the first neural network being adapted to a different task by the second neural network each time that new second input data is received by the second neural network as explained above and shown in the detailed rejection below and Examiner disagrees that this concept is neither taught nor suggested by the combinations utilized by the office.
Regarding Applicant’s arguments, see pages 6 and 7, Applicant argues that by way of example, Applicant's "second input data" cannot be understood as a "validation set" as alleged by the office. Office Action, p. 7. Such an interpretation is contrary to the broadest reasonable interpretation based on the specification. In re Suitco Surface, Inc., 603 F.3d 1255, 1260 (Fed. Cir. 2010) (stating "this court has instructed that any such construction be consistent with the specification ... and that claim language should be read in light of the specification as it would be interpreted by one of ordinary skill in the art"). A validation set is hold out data used in evaluation of a trained model. Le at p. 3, lines 30-34 ("training data 102 for training a neural network to perform a particular task and a validation set 104 for evaluating the performance of the neural network on the particular task"). Applicant submits that validation and training data are distinguishable from the claimed "second input data" being used to adapt "the first neural network." Claim 1; Specification at [0025].
Examiner’s Response: Applicant’s arguments have been fully considered but are not persuasive because firstly, the term “input data” is very broad and so general that under broadest reasonable interpretation, input data can be any data that is being fed to the system. Secondly, although Applicant argues that “second input data” cannot be understood as a “validation set” as such an interpretation is contrary to the broadest reasonable interpretation based on the specification this argument is incorrect because broadest reasonable interpretation applies to the claim language and not the art or specification etc. Thirdly, although Applicant argues that claim language should be read in light of the specification, Applicant appears to be reading the claim language in light of specification of reference Le which is incorrect as the phrase “claim language should be read in light of the specification “ refers to reading claim language in the light of the instant application’s specification and not in light of the specification of the prior art references. The claim language is mapped to references for teaching the claim limitations. Lastly, although Applicant submits that validation and training data are distinguishable from the claimed "second input data" being used to adapt "the first neural network." Applicant fails to provide persuasive reasons and explanation as to why.
Hence, the cited references teach the claims as explained above and shown in the detailed rejection below.
Regarding Applicant’s arguments, see page 7 (paragraph ), Applicant argues that to expedite prosecution, Applicant has amended the independent claims to recite, inter alia, that "the second neural network is adapted to generate the at least one task-specific parameter further based on third input data that provides information on the first input data; each of the second input data and the third input data being initially separately processed and combined into a combined input used to calculate the at least one task specific parameter." Claim 1. No new matter is entered. Specification at [0110]. Applicant therefore requests reconsideration and withdrawal of the rejections based on the interpretation that Le's "validation set" may be construed or understood as Applicant's "second input data" or as inherently comprising the "third input data" as claimed. Claims 1-2 (as previously presented).
Examiner’s Response: Applicant’s arguments have been fully considered but are not persuasive because the first part of this amendment has come from canceled claim 2 and Le teaches that as shown in the detailed rejection below. The second part of the amendment is new and upon further search and consideration Le teaches that as well as shown in the detailed rejection below. Le states on Page 4, Lines 14-22 that as another example, the system 100 can receive an input from a user specifying which data that is already maintained by the system 100 should be used for training the neural network, and then divide the specified data into the training data 102 and the validation set 104, and Examiner notes again that because “input data” is a broad term, “an input from a user” corresponds to the third input data as it provides information on the first input data which is just any kind of digital data and validation data corresponds to second input data as explained above. Hence, the cited references teach the newly added amendments.
Regarding Applicant’s arguments, see page 7 (paragraph ), Applicant argues that Applicant also continues to disagree with the Office's proposed combination of Le's teachings with the references of record. In each case the Office alleges a combination that relies on a feedback mechanism. Office Action, pp. 10, 15-17. However, none of the cited documents, even considered in combination, teach or suggest processing the second and third input data, and further combining the resulting processed outputs in generating a task specific parameter by operation of the second network, as claimed. Claim 1 (as amended). Accordingly, Applicant requests reconsideration and withdrawal of the rejections of the claims.
Examiner’s Response: Applicant’s arguments have been fully considered but are not persuasive because Le teaches the newly added limitation “each of the second input data and the third input data being initially separately processed and combined into a combined input used to calculate the at least one task specific parameter” at least in Pages 4, 10, and 22. Page 4 shows that the system can receive an input from a user separating the training data and validation data showing that input data being initially separated. Page 10 states that each node in the tree merges two inputs to generate an output showing that input data are combined into a combined input used to calculate at least one task specific parameter. Hence, Le still teaches the amended claim 1 (and similarly claim 10) as explained above and shown in the detailed rejection below.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
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.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
Claims 1, 3-6, 10, 12-16, 19, 21, and 22 are rejected under 35 U.S.C. 103 as being unpatentable over Le (WO2018081563A1) in view of Thaler (US 6018727 A).
Regarding claim 1:
Le teaches: an adaptable neural network system for processing first input data, the neural network system comprising (Pg. 2 Lines 20-23 “a system implemented as computer programs on one or more computers in one or more locations can determine, using a controller neural network, an architecture for a child neural network that is configured to perform a particular neural network task.”)”
a non-transitory computer readable storage medium comprising computer readable program instructions for execution by a processor to provide ([Page 14, Lines 19-22] Embodiments of the subject matter described in this specification can be implemented as one or more computer programs, i.e., one or more modules of computer program instructions encoded on a tangible non transitory storage medium for execution by, or to control the operation of, data processing apparatus):
a first neural network formed of a plurality of sequential layers, wherein the first neural network is adapted to generate output data by processing first input data using the plurality of sequential layers (Pg. 5 Lines 4-8 “In particular, during an iteration of the training procedure, the system generates a batch of sequences using the controller neural network in accordance with current values of the controller parameters. For each output sequence in the batch, the training engine trains an instance of the child neural network that has the architecture defined by the output sequence on the training data and evaluates the performance of the trained instance on the validation set.” Pg. 2 Lines 24-26 “The child neural network can be configured to receive any kind of digital data input and to generate any kind of score, classification, or regression output based on the input.” *Note: The “first neural network” of the instant application corresponds to the “child neural network” of the prior art citation, and the “first input data” corresponds to the “digital data input” of the second citation.);
a second neural network adapted to generate at least one task-specific parameter for the plurality of sequential layers of the first neural network based on second input data indicative of a desired task to be performed on the first input data (Pg. 5 Lines 4-13 “In particular, during an iteration of the training procedure, the system generates a batch of sequences using the controller neural network in accordance with current values of the controller parameters. For each output sequence in the batch, the training engine trains an instance of the child neural network that has the architecture defined by the output sequence on the training data and evaluates the performance of the trained instance on the validation set. The controller parameter updating engine then uses the results of the evaluations for the output sequences in the batch to update the current values of the controller parameters to improve the expected performance of the architectures defined by the output sequences generated by the controller neural network on the task.” *Note: The “second neural network” corresponds to the “controller neural network” of the citation. The “second input indicative of a desired task to be performed” corresponds to the “validation set” in the prior art. The controller neural network generates an output sequence that defines the architecture of the child network. The controller neural network generates these outputs based on the evaluation of the child network against the validation set. The validation set is being used to score the child network on a particular task, and thus is inherently indicative of the desired task to be performed.);
wherein the second neural network is adapted to generate the at least one task- specific parameter further based on third input data that provides information on the first input data ([Page 4, Lines 14-22] As another example, the system 100 can receive an input from a user specifying which data that is already maintained by the system 100 should be used for training the neural network, and then divide the specified data into the training data 102 and the validation set 104. The neural architecture search system 100 includes a controller neural network 110, a training engine 120, and a controller parameter updating engine 130. The controller neural network 1 10 is a neural network that has parameters, referred to in this specification as "controller parameters," and that is configured to generate output sequences in accordance with the controller parameters. Note: Controller neural network corresponds to second neural network. an input from a user corresponds to the third input data as it provides information on the first input data which is just any kind of digital data);
each of the second input data and the third input data being initially separately processed and combined into a combined input used to calculate the at least one task specific parameter ([Page 4, Lines 14-17] As another example, the system 100 can receive an input from a user specifying which data that is already maintained by the system 100 should be used for training the neural network, and then divide the specified data into the training data 102 and the validation set 104. [Page 10, Lines 30-33] Each node in the tree merges two inputs to generate an output, and, for each node, the output sequence includes: (i) data identifying a combination method to combine the two inputs and (ii) an activation function to be applied to the combination of the two inputs to generate the output. [Page 22, Lines 1-4] The method of claim 14, wherein each node in the tree merges two inputs to generate an output, and wherein, for each node, the output sequence includes data identifying a combination method to combine the two inputs and an activation function to be applied to the combination of the two inputs to generate the output).
However, Le does not explicitly disclose: and a neural network modifier adapted to modify the first neural network based on the at least one task-specific parameter generated by the second neural network, to thereby adapt the first neural network to the desired task, the first neural network being adapted to a different task by the second neural network each time that new second input data is received by the second neural network.
Thaler teaches, in an analogous system: and a neural network modifier adapted to modify the first neural network based on the at least one task-specific parameter generated by the second neural network, to thereby adapt the first neural network to the desired task, the first neural network being adapted to a different task by the second neural network each time that new second input data is received by the second neural network ([Abstract] a first neural network trained to produce input-output maps within a predetermined initial knowledge domain, an apparatus for subjecting the neural network to perturbations which may produce changes in the predetermined knowledge domain, the neural network having an optional output for feeding the outputs of the first neural network to a second neural network that evaluates the outputs based on training within the second neural network. The device may also include a reciprocal feed back connection from the output of the second neural network to the first neural network to further influence and change what takes place in the aforesaid neural network. [Column 1, Lines 51-58] Instead, information emerges spontaneously as a result of any number of stochastic and/or systematic processes applied to the characterizing parameters of the networks involved. With this tandem arrangement of the free-running neural network and its policing counterpart, it is possible to generate a notion that is superior in quality to anything generated by a known device or machine similarly exposed or perturbed. Note: Change what takes place in the aforesaid neural network corresponds to the first neural network being adapted to a different task by the second neural network).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the adaptable neural network system of Le to incorporate the teachings of Thaler to use a neural network modifier adapted to modify the first neural network based on the at least one task-specific parameter generated by the second neural network, to thereby adapt the first neural network to the desired task, the first neural network being adapted to a different task by the second neural network each time that new second input data is received by the second neural network. One would have been motivated to do this modification because doing so would give the benefit of generating a notion that is superior in quality to anything generated by a known device or machine similarly exposed or perturbed as taught by Thaler [Column 1, Lines 51-58].
Regarding claim 3:
The system of Le and Thaler teaches: The adaptable neural network system of claim 1 (as shown above).
Le further teaches: wherein the second input data comprises an input query for the adaptable neural network system, indicative of a desired task to be performed (Pg. 4 Lines 4-6 “Generally, the training data and the validation set both include a set of neural network inputs and, for each network input, a respective target output that should be generated by the child neural network to perform the particular task.” *Note: The second input has been previously mapped to the validation set which is comprised of a set of target outputs. This set of target outputs corresponds to the “input query” of the claim. These target outputs are indicative of the desired task to be performed as they are the desired outputs of the task.).
Regarding claim 4:
The system of Le and Thaler teaches: The adaptable neural network system of claim 3 (as shown above).
Le further teaches: wherein the input query comprises a hypothesized property associated with the first input data, and the first neural network is adapted to determine a value indicating whether the hypothesized property is correct (Pg. 2 Lines 27-31 “For example, if the inputs to the child neural network are images or features that have been extracted from images, the output generated by the child neural network for a given image may be scores for each of a set of object categories, with each score representing an estimated likelihood that the image contains an image of an object belonging to the category.” *Note: The input query has been established as the set of target outputs in the validation set. In this example, the set of target outputs would be the categories for the images being input. Any category in particular can serve as the hypothesized property, and the child neural network is adapted to generate a value – referred to as a score in the prior art – representing whether the image has this property and thus belongs in the category.).
Regarding claim 5:
The system of Le and Thaler teaches: The adaptable neural network system of claim 1 (as shown above).
Le further teaches: wherein the plurality of sequential layers of the first network comprises a predetermined number of adjustable parameters, and the second neural network is adapted to generate a task-specific parameter for each of the predetermined number of adjustable parameters (Pg. 8 Lines 3-5 “In the example of FIG. 2A, the child neural network is a convolutional neural network and the hyperparameters include hyperparameters for each convolutional neural network layer in the child neural network.” Pg. 8 Lines 12-15 “In the example of FIG. 2A, for a convolutional layer, the hyperparameters that define the operations performed by the layer are the number of filters of the layer, the filter height for each filter, the filter width for each filter, the stride height for applying each filter, and the stride width for each filter.” Pg. 7 Lines 18-22 “Thus, to generate a hyperparameter value for a given time step in an output sequence, the system provides as input to the controller neural network the value of the hyperparameter at the preceding time step in the output sequence and the controller neural network generates an output for the time step that defines a score distribution over possible values of the hyperparameter at the time step.” *Note: The “adjustable parameters” of the claim correspond to the hyperparameters of the prior art. It can be seen in the second citation that one embodiment of the prior art has a predetermined number of hyperparameters – which are listed. The last citation shows that the controller neural network is generating a “task-specific parameter” – in this case a score distribution – for each hyperparameter).
Regarding claim 6:
The system of Le and Thaler teaches: The adaptable neural network system of claim 1 (as shown above).
Le further teaches: wherein the second neural network comprises a plurality of sequential layers and the at least one task-specific parameter is output by a sequentially last of the plurality of sequential layers (Pg. 6 Lines 30-31 “The controller neural network is a recurrent neural network that includes one or more recurrent neural network layers” Pg. 7 Lines 6-11 “The controller neural network also includes a respective output layer for each time step in the output sequence, e.g., output layers for time steps, respectively. Each of the output layers is configured to receive an output layer input that includes the updated hidden state at the time step and to generate an output for the time step that defines a score distribution over possible values of the hyperparameter at the time step.” *Note: The “output layer” is the last layer in the controller neural network, and it outputs a “score distribution” which is the task-specific parameter).
Regarding claims 10 and 12-14:
They recite limitations similar to claims 1, 3, 5, and 6 and are rejected on the same basis.
Regarding claim 15:
The system of Le and Thaler teaches: A non-transitory computer readable storage medium comprising program code for implementing the method of claim 10 (as shown above).
Le further teaches: when said program code is run on a computer ([Page 14, Lines 19-22] Embodiments of the subject matter described in this specification can be implemented as one or more computer programs, i.e., one or more modules of computer program instructions encoded on a tangible non transitory storage medium for execution by, or to control the operation of, data processing apparatus).
Regarding claim 16:
The system of Le and Thaler teaches: The method of claim 10 (as shown above).
Le further teaches: wherein the least one task-specific parameter adapts the first neural network to a new task ([Page 1, Lines 28-31] The system can effectively determine novel neural network architectures that are adapted for a particular task, allowing the resulting child neural network to have an improved performance on the task).
Regarding claim 19:
The system of Le and Thaler teaches: The method of claim 10 (as shown above).
However, Le does not explicitly disclose: wherein the first neural network is a generic neural network adapted by the at least one task-specific parameter.
Thaler teaches, in an analogous system: wherein the first neural network is a generic neural network adapted by the at least one task-specific parameter ([Abstract] a first neural network trained to produce input-output maps within a predetermined initial knowledge domain, an apparatus for subjecting the neural network to perturbations which may produce changes in the predetermined knowledge domain, the neural network having an optional output for feeding the outputs of the first neural network to a second neural network that evaluates the outputs based on training within the second neural network. The device may also include a reciprocal feed back connection from the output of the second neural network to the first neural network to further influence and change what takes place in the aforesaid neural network. [Column 1, Lines 51-58] Instead, information emerges spontaneously as a result of any number of stochastic and/or systematic processes applied to the characterizing parameters of the networks involved. With this tandem arrangement of the free-running neural network and its policing counterpart, it is possible to generate a notion that is superior in quality to anything generated by a known device or machine similarly exposed or perturbed. [Column 12, Lines 27-40] a first artificial neural network portion, said first artificial neural network having been previously trained in a predefined field of endeavor and having an input portion, an output portion, and a particular knowledge domain as established therein for the predefined field of endeavor through prior training of said first neural network portion, which first trained neural network portion would be normally operable in accordance with its basic design constraints and the established knowledge domain to produce established outputs reflective of standard design concepts in the predefined field of endeavor in response to a pattern of inputs supplied to said first neural network portion at the input portion thereof. Note: Change what takes place in the aforesaid neural network corresponds to the first neural network being adapted by the at least one task-specific parameter. Standard corresponds generic).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the method of Le to incorporate the teachings of Thaler wherein the first neural network is a generic neural network adapted by the at least one task-specific parameter. One would have been motivated to do this modification because doing so would give the benefit of generating a notion that is superior in quality to anything generated by a known device or machine similarly exposed or perturbed as taught by Thaler [Column 1, Lines 51-58].
Regarding claim 21:
The system of Le and Thaler teaches: The method of claim 19 (as shown above).
Le further teaches: wherein the second input is received from a user and comprises a new classification task for which the first neural network is not yet adapted ([Page 2, Lines 24-26] The child neural network can be configured to receive any kind of digital data input and to generate any kind of score, classification, or regression output based on the input. [Page 4, Lines 9-10] The system 100 can receive the training data 102 and the validation set 104 in any of a variety of ways. [Page 4, Lines 14-33] As another example, the system 100 can receive an input from a user specifying which data that is already maintained by the system 100 should be used for training the neural network, and then divide the specified data into the training data 102 and the validation set 104. The neural architecture search system 100 includes a controller neural network 110, a training engine 120, and a controller parameter updating engine 130. The controller neural network 1 10 is a neural network that has parameters, referred to in this specification as "controller parameters," and that is configured to generate output sequences in accordance with the controller parameters. Each output sequence generated by the controller neural network 110 defines a respective possible architecture for the child neural network. In particular, each output sequence includes a respective output at each of multiple time steps and each time step in the output sequence corresponds to a different hyperparameter of the architecture of the child neural network. Thus, each output sequence includes, at each time step, a respective value of the corresponding hyperparameter. Collectively, the values of the hyperparameters in a given output sequence define an architecture for the child neural network. Generally, a hyperparameter is a value that is set prior to the commencement of the training of the child neural network and that impacts the operations performed by the child neural network. [Page 12, Lines 32-33] The system can train each child neural network for a specified amount of time or a specified number of training iterations. [Page 13, Lines 6 and 7] For example, the accuracy can be a perplexity measure when the outputs are sequences or a classification error rate when the task is a classification task. Note: Validation set corresponds to second input data and shows that the system receives it as an input from the user. Child neural network corresponds to the first neural network and a hyperparameter is a value that is set prior to the commencement of the training of the child neural network shows that the first neural network is not yet adapted).
Regarding claim 22:
The system of Le and Thaler teaches: The method of claim 10 (as shown above).
However, Le does not explicitly disclose: wherein the at least one task-specific parameter comprises one or more of weights and biases for the first neural network.
Thaler teaches, in an analogous system: wherein the at least one task-specific parameter comprises one or more of weights and biases for the first neural network ([Column 6, Lines 53-58] The first IE network is trained to produce outputs within the knowledge domain of its training. The introduction of perturbations to any number of ANN features cause the IE to wander through the knowledge domain producing meaningful outputs under the constraints of its connection strengths and biases. [Column 13, Lines 60-63] wherein the first artificial neural network portion includes a plurality of weights, biases, and activation levels associated therewith).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the method of Le to incorporate the teachings of Thaler wherein the at least one task-specific parameter comprises one or more of weights and biases for the first neural network. One would have been motivated to do this modification because doing so would give the benefit of causing the IE to wander through the knowledge domain producing meaningful outputs under the constraints of its connection strengths and biases as taught by Thaler [Column 6, Lines 56-58].
Claims 7-9 are rejected under 35 U.S.C. 103 as being unpatentable over Le (WO2018081563A1) in view of Thaler (US 6018727 A) and further in view of Hasan (US20190221312A1).
Regarding claim 7:
The system of Le and Thaler teaches: The adaptable neural network system of claim 1 (as shown above).
However, Le fails to teach: wherein the first input data comprises clinical information of a subject and the second input data comprises a hypothesized diagnosis or symptom of the subject.
Hasan teaches, in an analogous system: wherein the first input data comprises clinical information of a subject and the second input data comprises a hypothesized diagnosis or symptom of the subject (Pg. 5 Par. 42 “The first electronic health record can be an electronic document that includes headings and text (e.g., sentences) that identify the patient diagnosis, as well as information related to the patient diagnosis such as, for example, symptoms and test results” Pg. 5 Par. 43 “The method can further include a block of accessing a first image associated with the patient diagnosis. Each of the first electronic health record and the first image can correspond to a patient, or multiple patients, that received the patient diagnosis.”).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the invention to utilize the teachings of Hasan to modify the teachings of Le and Thaler to apply the neural network system in a medical context by having the first input comprise a medical image and the second input comprise diagnosis/symptom information. The motivation for doing so would have been to assist with identifying a more accurate diagnosis for patients (Hasan Pg. 2 Par. 18).
Regarding claim 8:
The system of Le, Thaler, and Hasan teaches: The adaptable neural network system of claim 7 (as shown above).
However, the system of Le and Thaler fails to teach: wherein the first input data comprises a medical image of the subject.
Hasan further teaches, in an analogous system: wherein the first input data comprises a medical image of the subject (Pg. 5 Par. 43 “accessing a first image associated with the patient diagnosis” *Note: The same argument for obviousness for claim 7 applies to claim 8.).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the invention to modify the combined teachings of Le and Thaler to incorporate the teachings of Hasan wherein the first input data comprises a medical image of the subject. The motivation for doing so would have been to assist with identifying a more accurate diagnosis for patients (Hasan Pg. 2 Par. 18).
Regarding claim 9:
The system of Le, Thaler, and Hasan teaches: The adaptable neural network system of claim 8 (as shown above).
However, Le fails to teach: wherein the second neural network is further adapted to generate the at least one task-specific parameter based on metadata associated with the medical image of the subject.
Hasan further teaches, in an analogous system: wherein the second neural network is further adapted to generate the at least one task-specific parameter based on metadata associated with the medical image of the subject (Pg. 4 Par. 43 “the first image can be associated with metadata that can provide descriptions about different portions of the image, and/or identify the clinician, patient, and/or any other data relevant to the patient diagnosis”).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the invention to utilize the teachings of Hasan to modify the teachings of Le and Thaler to use the image meta data as a third input. The motivation for doing so would be to provide additional data to the neural network system – such as a description of the image – that may be relevant to the patient diagnosis (Hasan Pg. 4 Par. 43).
Claim 17 is rejected under 35 U.S.C. 103 as being unpatentable over Le (WO2018081563A1) in view of Thaler (US 6018727 A) and further in view of Bennett et al (US 6615172 B1).
Regarding claim 17:
The system of Le and Thaler teaches: The method of claim 16 (as shown above).
However, the system of Le and Thaler fails to teach: wherein the second input data is structure-less text that is converted into a structured input query.
Bennett teaches, in an analogous system: wherein the second input data is structure-less text that is converted into a structured input query ([Column 7, Lines 21-23] After the user's question is decoded by the speech recognition engine (SRE) located at the server, the question is converted to a structured query language (SQL) query. [Column 38, Lines 53-55] a query formulation engine adapted to convert said recognized words and said word phrases into a structured query).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the method of Le and Thaler to incorporate the teachings of Bennett wherein the second input data is structure-less text that is converted into a structured input query. One would have been motivated to do this modification because doing so would give the benefit of presenting this query to the software for preliminary processing as taught by Bennett [Column 7, Lines 24-26].
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Zoph et al (2017) discloses NEURAL ARCHITECTURE SEARCH WITH REINFORCEMENT LEARNING.
Thaler (US 5659666 A) discloses Device For The Autonomous Generation Of Useful Information.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to CHAITANYA RAMESH JAYAKUMAR whose telephone number is (571)272-3369. The examiner can normally be reached Mon-Fri 9am-1pm.
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 Rivas can be reached on (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.
/C.R.J./Examiner, Art Unit 2128
/OMAR F FERNANDEZ RIVAS/Supervisory Patent Examiner, Art Unit 2128