DETAILED ACTION
The action is in response to the original filing on February 7, 2023 and the Remarks and Amendments filed on April 6, 2026. Claims 1-20 are pending and have been considered below. Claims 1, 11, and 20 are independent claims. Claims 1, 5-6, 8, 11, 15-16, 18, and 20 are amended. Claims 7 and 17 are canceled.
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 .
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.
Claims 1-2, 6, 11-12, 16, and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Wu et al. (US 20230153658 A1, hereinafter Wu) in view of Giri et al. (US 12014748 B1, hereinafter Giri) and further in view of Shu et al. (English Translation of CN 114580634 A, hereinafter Shu).
Regarding claim 1:
Regarding the limitation a method of providing automated explanations for inference services based on artificial intelligence using a cloud, the method comprising: requesting an inference response message to an inference container according to an inference service based on an inference request message received from a client, Wu teaches a method of providing automated explanations for inference services based on artificial intelligence (¶6 “a method, apparatus and system to automatically generate explanations for decision predictions that are generated by a machine learning algorithm”) using a cloud (Fig. 1 – 104, 108, Fig. 2 – 208, ¶69 “the network interface 208 can be configured to include or comprise a medium through which the machine learning module 104, the prediction explanation module 108, as well as other connected devices, can communicate with each other. The communication network… may be a… wireless communication network. Examples of suitable communication networks can include… a cloud network”) and an inference container according to an inference service (Fig. 1 – 102-106, ¶35 “The machine learning module 104, which in one embodiment comprises a machine learning algorithm, is generally configured to receive input data 102, referred to in this example as input data 1 to input data n. The machine learning module 104 will calculate an output 106 based on the input 102. In the examples herein, the output 106 is referred to as a decision prediction and referenced in FIG. 1 as Output 1 to Output n. As will be generally understood, the output 106 will generally comprise a decision prediction of the algorithm running in the machine learning module 104, based on the input data 102,” wherein a “machine learning module” that provides “decision predictions” functions in substantially the same way as an inference container according to an inference service). However, Wu fails to teach the method comprising: requesting an inference response message to an inference container according to an inference service based on an inference request message received from a client.
Giri, in the same field of endeavor, teaches the method comprising: requesting an inference response message to a hosting service… based on an inference request message received from a client (Fig. 1 – 130, 134, 136, 138, 140A-B, 160A-B, Col. 7, Lines 21-28 “a hosting service 134 of a machine learning service 130 to deploy a model as a hosted model 136 in association with an endpoint 138 that can receive inference requests from client applications 140A and/or 140B… provide the inference requests 160A to the associated hosted model(s) 136, and provide inference results 160B (e.g., predicted intent, predicted action, etc.) back to applications 140A and/or 140B”).
Regarding the limitation sending the inference request message and the inference response message to an imitation learning container linked with the inference container according to a mirroring setting, Wu teaches sending an input and an output to an imitation learning container linked with the inference container according to a mirroring setting (Fig. 1 – 102-110, ¶35 “the output 106 will generally comprise a decision prediction of the algorithm running in the machine learning module 104, based on the input data 102,” ¶36 “The prediction explanation module 108 is generally configured to access the input 102, the output 106 and generate additional information that reveals one or more causal relationships between the output 106 and the input 102,” wherein a “prediction explanation module” functions in substantially the same way as an imitation learning container; Fig. 1 depicts how the “prediction explanation module” can “access the input” and “the output” of the “machine learning module,” which functions in substantially the same way as according to a mirroring setting). However, Wu fails to teach the inference request message and the inference response message.
Giri teaches the inference request message (Fig. 1 – 130, 134, 138, 140A-B, Col. 7, Lines 21-24 “inference requests from client applications”) and the inference response message (Fig. 1 – 136, 140A-B, 160A-B, Col. 7, Lines 25-28 “provide inference results… back to applications”).
Regarding the limitation creating interpretation information of the inference container based on the inference request message and the inference response message, and providing the created interpretation information to the client, Wu teaches creating interpretation information of the inference container based on input and output (Fig. 1 – 100-110, ¶36 “The prediction explanation module 108 is generally configured to access the input 102, the output 106 and generate additional information… The explanation output 110 is the additional information from the prediction explanation module 108 and identifies a causal or structural relationship in the input data 102 to explain the reasoning for the output 106. The explanation output 110 is presented in a human understandable manner,” wherein “explanation output” encompasses interpretation information), and providing the created interpretation information to the client (Fig. 3 – 304-308, ¶42 “relationships between the input data and the output, the prediction data, are identified 306. The relationship data can be presented 308 to the user in a human understandable manner. For example, in one embodiment, an explanation of the reasoning to arrive at the prediction decision 106 can be presented, such as on a user interface of a computing device”). However, Wu fails to teach the inference request message and the inference response message.
Giri teaches the inference request message (Fig. 1 – 130, 134, 138, 140A-B, Col. 7, Lines 21-24) and the inference response message (Fig. 1 – 136, 140A-B, 160A-B, Col. 7, Lines 25-28).
Regarding the limitation wherein the creating of the interpretation information and the providing of the created interpretation information to the client includes providing an imitation inference result of the imitation learning container and the inference response message to the client along with the interpretation information when a critical condition is satisfied, Wu teaches wherein the creating of the interpretation information (Fig. 1 – 100-110, ¶36) and the providing of the created interpretation information to the client (Fig. 3 – 304-308, ¶42) includes providing an imitation inference result of the imitation learning container (Fig. 1 – 100-110, Fig. 2 – 202-204, ¶38 “the machine learning module 104 will generally comprise or otherwise be coupled to the first processor 202,” ¶39 “the prediction explanation module 108 includes or is coupled to the second processor 204. For example, in one embodiment, the second processor 204 is configured to run the algorithm(s) of the prediction explanation module 108 by accessing at least the output 106 of the machine learning module 104, as well as the inputs 102 and generate the explanation output 110,” ¶63 “the casual relationships generated by the second processor 204 are based on imitative training of another machine learning algorithm with access to the machine learning algorithm in the first processor 202… the second processor 204 has learned to generate the same or very similar predictions compared to the machine learning algorithm 104 run by the first processor 202, given the same input 102,” ¶52 “a given number of images belong to different classes. The input images are segmented into groups of pixels, also referred to herein as “visual concepts.” The groups of pixels are input into… the model implemented by the prediction explanation module 108. The top “n” visual concepts can be selected based on the output of the machine/deep model by ranking the score each visual concept corresponds to. This ranking is presented by the explanation output 110,” ¶55 “the identified causal relationships can be a structural representation of different components sharing causal relationships with the input data 102, or the output 106. As an example, when differentiating between a fire engine and a car, the fire engine will have different components or parts than the car. These different parts or structural representations will provide different visual concepts… The spatial relationships between these visual concepts can be utilized to generate explanations,” wherein the “ranking” presented by “the explanation output” generated based on “imitative training of another machine learning algorithm” encompasses an imitation inference result) … to the client along with the interpretation information (Fig. 3 – 304-308, ¶42 “The relationship data can be presented 308 to the user in a human understandable manner… an explanation of the reasoning to arrive at the prediction decision 106 can be presented, such as on a user interface of a computing device”). However, Wu fails to teach wherein the creating of the interpretation information and the providing of the created interpretation information to the client includes providing an imitation inference result of the imitation learning container and the inference response message to the client along with the interpretation information when a critical condition is satisfied.
Giri teaches and the inference response message (Fig. 1 – 136, 140A-B, 160A-B, Col. 7, Lines 25-28 “provide inference results… back to applications”). However, Giri fails to teach when a critical condition is satisfied.
Shu, in the same field of endeavor, teaches when a critical condition is satisfied (¶74 “continuing to adjust the second model according to the current similarity parameter until the calculated similarity parameter meets the similarity parameter target value, and stopping adjusting”).
Wu further teaches the imitation inference result includes an inference derived from object information provided by the client (Fig. 1 – 104-108, 110, ¶61 “the identified casual relationships can be corrected, either automatically by another algorithm, or manually by a user… this would occur when the output 106 is an incorrect prediction, based on the input 104. The correction can then be used to improve the performance of the machine learning algorithm… a user identifies a visual concept in the representation of an input image that should not be associated with the class of interest… The user in this case can remove that visual concept from the representation,” ¶65 “concept graphs are used to represent different categories, adjustments can be made to existing concept graphs, by for example, removing or replacing some of the concepts or removing or editing the edges, to represent new objects… by editing the concept graph for bus, the user can create the concept graph for a fire engine. This way the user can define representations for new objects which can then be “distilled” and transferred to machine learning algorithm 104 so that the machine learning algorithm 104 learns to encode these objects,” ¶52 “The top “n” visual concepts can be selected… This ranking is presented by the explanation output 110,” ¶63 “the second processor 204 has learned to generate the same or very similar predictions compared to the machine learning algorithm 104 run by the first processor 202, given the same input 102,” wherein user-defined “representations for new objects” encompasses object information provided by the client; one of ordinary skill in the art would recognize that a “second processor” coupled to a “prediction explanation module” that is trained to “generate the same or very similar predictions” to the “machine learning module” after a “user” has defined “representations for new objects” for the “machine learning algorithm” implies that the imitation inference result is or includes an inference derived from object information provided by the client).
Regarding the limitation and the critical condition is set to a condition in which the imitation inference result of the imitation learning container has a similarity of a reference value or greater with respect to an inference result according to the inference response message, Wu teaches the imitation inference result of the imitation learning container (Fig. 1 – 108, Fig. 2 – 204, ¶52, ¶63) and an inference result according to input (Fig. 1 – 102-106, ¶35 “The machine learning module 104 will calculate an output 106 based on the input 102”). However, Wu fails to teach and the critical condition is set to a condition in which the imitation inference result of the imitation learning container has a similarity of a reference value or greater with respect to an inference result according to the inference response message.
Giri teaches the inference response message (Fig. 1 – 136, 140A-B, 160A-B, Col. 7, Lines 25-28). However, Giri fails to teach and the critical condition is set to a condition in which the imitation inference result of the imitation learning container has a similarity of a reference value or greater with respect to an inference result according to the inference response message.
Shu teaches and the critical condition is set to a condition in which a second model’s output has a similarity of a reference value or greater with respect to a first model’s output (Fig. 1 – S103 of English Translation, ¶63 “Taking the first target hidden layer as the last hidden layer of the first model and the second target hidden layer as the last hidden layer of the second model as an example, the feature vector of the output of the last hidden layer is regarded as a vector with angle and length features in space, and the similarity between the first feature vector and the second feature vector can be measured by comprehensively considering the length and the angle through the similarity parameter,” ¶64 “the similarity parameter can be identified by Tanimoto coefficient… T (a, B) represents the similarity parameter between the first feature vector a and the second feature vector B,” ¶74 “Performing a model distillation process based on the similarity parameter (expressed as a T coefficient) transfers the knowledge of the first model to the second model, essentially a process of adjusting the model parameters of the second model based on the T coefficient. In order to optimize the model distillation effect, it is necessary to make the feature vector output by the last hidden layer of the second model and the feature vector output by the last hidden layer of the first model most similar… iteration updating is performed on the second model based on the T coefficient calculated in real time, so that the T coefficient finally reaches a similarity target value, and the similarity target value is the most similar value of the first feature vector and the second feature vector. And if the similarity coefficient does not meet the similarity parameter target value, continuing to adjust the second model according to the current similarity parameter until the calculated similarity parameter meets the similarity parameter target value, and stopping adjusting”).
Wu, Giri, and Shu are analogous to the claimed invention as all are from the same field of endeavor of machine learning. Therefore, it would have been obvious to one of ordinary skill in the art, before the filing date of the claimed invention, to have combined the inference request and response messages of Giri and the critical condition of Shu with the methodology of Wu. The motivation to do so is “to enable rapid execution of tasks (for example, rapid initialization of ML scoring container(s) 850, rapid execution of code 856 in ML scoring container(s), etc.) in response to deployment and/or execution requests” (Giri, Fig. 8 – 850, 856, Col. 23, Lines 47-51) and to improve “convenience and accuracy of model processing” (Shu, ¶44).
Regarding claim 2, Wu in view of Giri and further in view of Shu teaches the method of claim 1 (and thus the rejection of claim 1 is incorporated).
Wu teaches wherein the inference container and the imitation learning container have independent learning models (Fig. 1 – 104, 108, Fig. 2 – 202, 204, ¶38 “the machine learning module 104 will generally comprise or otherwise be coupled to the first processor 202,” ¶39 “the prediction explanation module 108 includes or is coupled to the second processor 204,” wherein the “machine learning module” and the “prediction explanation module” being coupled to their own separate processors encompasses wherein the inference container and the imitation learning container have independent learning models), and the imitation learning container is trained to imitate an inference of the inference container (Fig. 1 – 102-104, Fig. 2 – 202-204, ¶63 “ the casual relationships generated by the second processor 204 are based on imitative training of another machine learning algorithm with access to the machine learning algorithm in the first processor 202… the second processor 204 has learned to generate the same or very similar predictions compared to the machine learning algorithm 104 run by the first processor 202, given the same input 102”).
Regarding claim 6, Wu in view of Giri and further in view of Shu teaches the method of claim 1 (and thus the rejection of claim 1 is incorporated).
Wu teaches wherein the interpretation information includes at least one of an input feature (Fig. 1 – 102, 106-110, ¶36 “The explanation output 110 is the additional information from the prediction explanation module 108 and identifies a causal or structural relationship in the input data 102 to explain the reasoning for the output 106,” Fig. 4 – 400, 402, ¶43 “the method 400 is directed to generating the additional information that identifies the one or more causal relationships between the prediction output 106 of the machine learning algorithm 104 and the input data 102. In one embodiment, primitive concepts in the input data 102 are identified 402. Primitive concepts can generally be considered visual concepts that are extracted from the input data”, wherein “primitive concept” encompasses input feature) that acts on an imitation inference result of the imitation learning container (Fig. 1 – 102, 108, ¶55 “the causal relationships that are identified by the prediction explanation module 108 are spatial correlations between visual concepts of the input data 102. The spatial correlations can be used to identify patterns or other important visual concepts or cues in the input data that are used to form the decisions,” wherein “decisions” formed from “causal relationships” identified by “the prediction explanation module” encompass an imitation inference result) and an importance of the input feature (Fig. 1 – 106, Fig. 4 – 406-410, ¶45 “Correlations are calculated 406 between the prediction output(s) 106 and each component in the representation and the calculated correlations are converted 408 to causal importance scores… the correlations between input concept representations and prediction output can be mathematically quantified to certain values, normalized between 0 to 1,” ¶46 “the converted causal importance scores are configured to be visualized 410 in a human-understandable manner… Representations and primitive concepts will correspond to each other (i.e., a one-to-one correspondence). After the casual importance scores of each concept representation are generated, those primitive concepts can be shown to the user… with decreasing/increasing casual importance scores”), wherein the importance of the input feature includes at least one of a weight of an object in an image or a node level located in a tree structure (Fig. 4 - 400, ¶43 “An image with a jeep car (object of interest) in the foreground, is classified as “jeep car” by the first processor. The primitive concepts can be groups of image pixels (part/component… for example, the jeep logo on the car, the wheels and/or the windows, among other aspects of the car,” ¶45 “each component in the representation and the calculated correlations are converted 408 to causal importance scores… normalized between 0 to 1,” ¶46 “those primitive concepts can be shown to the user via the user interface of the computing device, with decreasing/increasing casual importance scores,” ¶55 “the windows, doors and tires on a fire engine will have different sizes, shapes and spatial relationships relative to similar parts found on a car. The spatial relationships between these visual concepts can be utilized to generate explanations,” given its broadest reasonable interpretation, a weight can encompass any numerical score; hence, the importance of an object in an image, for example, a “logo” or “wheels” encompasses at least one of a weight of an object).
Regarding claim 11:
Wu teaches a platform device for providing automated explanations for inference services based on artificial intelligence (Fig. 1 – 100, ¶34 “an apparatus 100 for generating explanations for predictions generated by a machine learning model or algorithm is illustrated. In one embodiment, the apparatus 100 comprises for example, a computing device”) using a cloud (Fig. 1 – 104, 108, Fig. 2 – 208, ¶69), the platform device comprising: a transceiver configured to transmit and receive a signal (Fig. 2 – 208, ¶72 “The network interface 208 includes suitable logic, circuitry, and/or interfaces that is configured to communicate with one or more external devices… Examples of the network interface 208 may include… a radio frequency (RF) transceiver”).
Wu further teaches a processor configured to process the signal (Fig. 1 – 104, 108, Fig. 2 – 202-204, 208, ¶66 “the processors 202 and 204… generally include suitable logic, circuitry, interfaces and/or code that is configured to process data provided as an input… the processors 202 and 204 may be one or more individual processors,” ¶69 “the network interface 208 can be configured to include or comprise a medium through which the machine learning module 104, the prediction explanation module 108, as well as other connected devices, can communicate with each other. The communication network… may be a… wireless communication network,” ¶38-¶39, wherein “the network interface” that can communicate wirelessly with each module coupled to their respective processors implies the processors are configured to process the signal, for example, “a radio frequency” received from “the network interface”).
Regarding the limitation wherein the processor is configured to request an inference response message to an inference container according to an inference service based on an inference request message received from a client, Wu teaches an inference container according to an inference service (Fig. 1 – 102-106, ¶35, as explained above with respect to claim 1). However, Wu fails to teach wherein the processor is configured to request an inference response message to an inference container according to an inference service based on an inference request message received from a client.
Giri teaches wherein the processor is configured to (Col. 12, Lines 1-8 “Some or all of the operations… or other processes described herein… are performed under the control of one or more computer systems configured with executable instructions… executing collectively on one or more processors”) request an inference response message to a hosting service… based on an inference request message received from a client (Fig. 1 – 130, 134, 136, 138, 140A-B, 160A-B, Col. 7, Lines 21-28).
Regarding the limitation send the inference request message and the inference response message to an imitation learning container linked with the inference container according to a mirroring setting, Wu teaches send input and output to an imitation learning container linked with the inference container according to a mirroring setting (Fig. 1 – 102-110, ¶35-36 as explained above with respect to claim 1). However, Wu fails to teach the inference request message and the inference response message.
Giri teaches the inference request message (Fig. 1 – 130, 134, 138, 140A-B, Col. 7, Lines 21-24 “inference requests from client applications”) and the inference response message (Fig. 1 – 136, 140A-B, 160A-B, Col. 7, Lines 25-28 “provide inference results… back to applications”).
Regarding the limitation create interpretation information of the inference container and provide the created interpretation information to the client when the imitation learning container is trained based on the inference request message and the inference response message to satisfy a critical condition, Wu teaches create interpretation information of the inference container (Fig. 1 – 100-110, ¶36 as explained above with respect to claim 1) and provide the created interpretation information to the client when the imitation learning container is trained based on input and output (Fig. 1 – 102-108, Fig. 2 – 202, 204, ¶39, ¶63, Fig. 3 – 304-308, ¶42). However, Wu fails to teach create interpretation information of the inference container and provide the created interpretation information to the client when the imitation learning container is trained based on the inference request message and the inference response message to satisfy a critical condition.
Giri teaches the inference request message (Fig. 1 – 130, 134, 138, 140A-B, Col. 7, Lines 21-24) and the inference response message (Fig. 1 – 136, 140A-B, 160A-B, Col. 7, Lines 25-28). However, Giri fails to teach create interpretation information of the inference container and provide the created interpretation information to the client when the imitation learning container is trained based on the inference request message and the inference response message to satisfy a critical condition.
Shu teaches to satisfy a critical condition (¶74 “until the calculated similarity parameter meets the similarity parameter target value”).
Regarding the limitation wherein the creation of the interpretation information and the provision of the created interpretation information to the client includes, by the processor, providing an imitation inference result of the imitation learning container and the inference response message to the client along with the interpretation information when the critical condition is satisfied, Wu teaches wherein the creation of the interpretation information (Fig. 1 – 100-110, ¶36) and the provision of the created interpretation information to the client (Fig. 3 – 304-308, ¶42) includes, by the processor (Fig. 1 – 100, Fig. 2 – 202-204, ¶66 “The processors are configured to respond to and process instructions that drive the apparatus 100”), providing an imitation inference result of the imitation learning container (Fig. 1 – 100-110, Fig. 2 – 202-204, ¶38-39, ¶63, ¶52, ¶55 all as explained above with respect to claim 1) … to the client along with the interpretation information (Fig. 1 – 110, ¶36, Fig. 3 – 304-308, ¶42). However, Wu fails to teach wherein the creation of the interpretation information and the provision of the created interpretation information to the client includes, by the processor, providing an imitation inference result of the imitation learning container and the inference response message to the client along with the interpretation information when the critical condition is satisfied.
Giri teaches and the inference response message (Fig. 1 – 136, 140A-B, 160A-B, Col. 7, Lines 25-28). However, Giri fails to teach when the critical condition is satisfied.
Shu teaches when the critical condition is satisfied (¶74).
Wu further teaches the imitation inference result includes an inference derived from object information provided by the client (Fig. 1 – 104-108, 110, ¶61, ¶65, ¶52, ¶63 all as explained above with respect to claim 1).
Regarding the limitation and the critical condition is set to a condition in which the imitation inference result of the imitation learning container has a similarity of a reference value or greater with respect to an inference result according to the inference response message, Wu teaches the imitation inference result of the imitation learning container (Fig. 1 – 108, Fig. 2 – 204, ¶52, ¶63) and an inference result according to input (Fig. 1 – 102-106, ¶35). However, Wu fails to teach and the critical condition is set to a condition in which the imitation inference result of the imitation learning container has a similarity of a reference value or greater with respect to an inference result according to the inference response message.
Giri teaches the inference response message (Fig. 1 – 136, 140A-B, 160A-B, Col. 7, Lines 25-28). However, Giri fails to teach and the critical condition is set to a condition in which the imitation inference result of the imitation learning container has a similarity of a reference value or greater with respect to an inference result according to the inference response message.
Shu teaches and the critical condition is set to a condition in which a second model’s output has a similarity of a reference value or greater with respect to a first model’s output (Fig. 1 – S103, Fig. 1 – S103, ¶63-64, ¶74).
Wu, Giri, and Shu are analogous to the claimed invention as all are from the same field of endeavor of machine learning. Therefore, it would have been obvious to one of ordinary skill in the art, before the filing date of the claimed invention, to have combined the inference request and response messages of Giri and the critical condition of Shu with the methodology of Wu. The motivation to do so is “to enable rapid execution of tasks (for example, rapid initialization of ML scoring container(s) 850, rapid execution of code 856 in ML scoring container(s), etc.) in response to deployment and/or execution requests” (Giri, Fig. 8 – 850, 856, Col. 23, Lines 47-51) and to improve “convenience and accuracy of model processing” (Shu, ¶44).
Claims 12 and 16 are device claims that contains similar limitations to those of claims 2 and 6, respectively. Therefore, claims 12 and 16 are rejected under substantially the same rationale as claims 2 and 6, respectively.
Regarding claim 20:
Regarding the limitation a system for providing automated explanations for inference services based on artificial intelligence using a cloud, the system comprising: a client computer configured to request creation or use of an inference service, Wu teaches a system for providing automated explanations for inference services based on artificial intelligence (¶6) using a cloud (Fig. 1 – 104, 108, Fig. 2 – 208, ¶69), the system comprising: a client computer (¶46 “a user interface of an associated computing device”) and an inference service (Fig. 1 – 102-106, ¶35 “The machine learning module 104, which in one embodiment comprises a machine learning algorithm, is generally configured to receive input data 102… The machine learning module 104 will calculate an output 106 based on the input 102”). However, Wu fails to teach a system for providing automated explanations for inference services based on artificial intelligence using a cloud, the system comprising: a client computer configured to request creation or use of an inference service.
Giri teaches client applications configured to request creation or use of an inference service (Fig. 1 – 130, 134, 136, 138, 140A-B, 160A-B, Col. 7, Lines 21-28 “a hosting service 134 of a machine learning service 130 to deploy a model as a hosted model 136 in association with an endpoint 138 that can receive inference requests from client applications 140A and/or 140B… provide the inference requests 160A to the associated hosted model(s) 136, and provide inference results 160B (e.g., predicted intent, predicted action, etc.) back to applications 140A and/or 140B”).
Regarding the limitation a platform device for providing automated explanations for inference services including a processor that processes a request for the inference service, Wu teaches a platform device for providing automated explanations for inference services including a processor (Fig. 1 – 100, ¶34, Fig. 2 – 202-204, ¶66) and the inference service (Fig. 1 – 102-106, ¶35). However, Wu fails to teach a platform device for providing automated explanations for inference services including a processor that processes a request for the inference service.
Giri teaches a processor that processes a request for services (Fig. 11 – 1100, 1170, 1175, Col. 29, Lines 61-63 “a computer system 1100 includes one or more offload cards 1170 (including one or more processors 1175,” Col. 30, Lines 18-19 “the offload cards 1170 can accommodate requests from other entities,” Fig. 8 – 802, Col. 26, Lines 2-6 “the user devices 802 can execute a stand-alone application… for submitting training requests, deployment requests, and/or execution requests”).
Claim 20 is a device claim that contains similar limitations to those of claim 11. Therefore, claim 20 is rejected under substantially the same rationale as claim 11.
Claims 3, 4, 9, 13, 14 and 19 are rejected under U.S.C. 103 as being unpatentable over Wu in view of Giri and further in view of Shu, and further in view of Maccanti et al. (US 20230171164 A1, hereinafter Maccanti).
Regarding claim 3, Wu in view of Giri and further in view of Shu teaches the method of claim 1 (and thus the rejection of claim 1 is incorporated).
Regarding the limitation wherein the imitation learning container is created based on an inference service descriptor used to create the inference service, and the inference service descriptor includes state information including first access information and input/output specifications of the inference container, Wu teaches the imitation learning container (Fig. 1 – 108, ¶36), the inference service (Fig. 1 – 102-106, ¶35 “The machine learning module 104 will calculate an output 106 based on the input 102”), and first access information (Fig. 1 – 100-106, Fig. 2 – 208, ¶69 “the network interface 208 can be configured to include or comprise a medium through which the machine learning module 104, the prediction explanation module 108… can communicate with each other. The communication network… may be a… wireless communication network… The devices of the system or apparatus 100 are potentially configured to connect to the communication network, in accordance with various wired and wireless communication protocols. Examples of such… wireless communication protocols may include… Transmission Control Protocol and Internet Protocol (TCP/IP),” one of ordinary skill in the art would recognize that an IP address, or first access information, is implicit in order for the “machine learning module,” or inference container, to connect to a “communication network” in accordance with “Internet Protocol”) and input/output specifications of the inference container (¶51-53 “the causal relationships that are identified by the prediction explanation module 108 are spatial correlations between visual concepts of the input data 102… a given number of images belong to different classes… the identification of visual patterns can be used to explain why an input image is a cat, or why an input image is not a dog,” ¶60 “if the top identifier of fire truck is the wheels, and an image of a truck with similar wheels is the input 102, the output 106 will likely predict the input image 102 as a fire engine,” wherein input/output specifications, when given its broadest reasonable interpretation, includes the type of input, for example, written, numerical, or visual data that the inference container is configured to receive; in one embodiment, the “Input Data” for the “Machine Learning Module” as illustrated in Figure 1 is described as “input images”). However, Wu fails to teach wherein the imitation learning container is created based on an inference service descriptor used to create the inference service, and the inference service descriptor includes state information including first access information and input/output specifications of the inference container.
Maccanti, in the same field of endeavor, teaches wherein a container is created based on an inference service descriptor (Fig. 5 – 500, ¶66 “the system generates… a container that comprises a model server associated with the requested endpoint and an extension… the model server refers to code and/or configuration data for implementing the model server,” wherein “the requested endpoint and an extension” encompass an inference service descriptor) used to create access to an inference service (¶63 “system receives a request to create an endpoint to provide machine learning services… endpoint creation refers to the system enabling itself to receive requests to interact with a machine learning model… the endpoint is associated with a network address, to which requests to access the machine learning model are directed… model servers associated with the endpoint are associated with network addresses”), and the inference service descriptor includes state information including instructions for interfacing with an inference service (¶63 “a model server associated with the requested endpoint and an extension… the extension refers to code that at least includes instructions for interfacing with the model server,” Fig. 3 – 312-316, 340-342, ¶51 “the extension 312 interfacing with the model server 314 to provide it with parameters 340, metadata 342, or other information to be used by the machine learning model 316,” wherein an “extension” that provides “parameters” and “metadata” to be used by a “machine learning model” encompasses state information).
Wu and Maccanti are analogous to the claimed invention as they are all from the same field of endeavor of machine learning. Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing data of the claimed invention, to combine the inference service descriptor and state information of Maccanti with the methodology of Wu. The motivation to do so is to be able to handle a surge in demand for machine learning services (Maccanti, ¶14-15 “Using a dedicated server may not work well if there is a surge in demand because the capacity of the dedicated server instance may be limited… this approach would typically require… administrative burden… To address these issues, the user is able to request that access to a machine learning model be provided using a serverless configuration”).
Regarding claim 4, Wu in view of Giri and further in view of Shu, and further in view of Maccanti teaches the method of claim 3 (and thus the rejection of claim 3 is incorporated).
Regarding the limitation wherein the mirroring setting is performed by an ingress setting for mirroring the inference response message to the imitation learning container together with the sent inference request message based on the first access information, Wu wherein the mirroring setting is performed by an ingress setting for mirroring an input to the imitation learning container together with an output (Fig. 1 – 102-108, Fig. 2 – 204, ¶39 “the prediction explanation module 108 includes or is coupled to the second processor 204. For example, in one embodiment, the second processor 204 is configured to run the algorithm(s) of the prediction explanation module 108 by accessing at least the output 106 of the machine learning module 104, as well as the inputs 102,” wherein “is configured” implies a configuration, which encompasses a setting; one of ordinary skill in the art would recognize that a setting of the "the second processor” to “run the algorithm(s) of the prediction explanation module by accessing at least the output of the machine learning module, as well as the inputs” functions in substantially the same way as an ingress setting for mirroring the “output” to the imitation learning container along with the “input”) based on the first access information (Fig. 1 – 100-106, Fig. 2 – 208, ¶69 “the network interface 208 can be configured to include or comprise a medium through which the machine learning module 104, the prediction explanation module 108… can communicate with each other. The communication network… may be a… wireless communication network… The devices of the system or apparatus 100 are potentially configured to connect to the communication network, in accordance with various wired and wireless communication protocols. Examples of such… wireless communication protocols may include… Transmission Control Protocol and Internet Protocol (TCP/IP),” wherein a “network interface” that allows the “machine learning module” and “the prediction explanation module” to communicate wirelessly with each other in accordance with “Internet Protocol” implies communication through IP addresses or based on the first access information). However, Wu fails to teach wherein the mirroring setting is performed by an ingress setting for mirroring the inference response message to the imitation learning container together with the sent inference request message based on the first access information.
Giri teaches the inference response message (Fig. 1 – 136, 140A-B, 160A-B, Col. 7, Lines 25-28) and the sent inference request message (Fig. 1 – 130, 134, 138, 140A-B, Col. 7, Lines 21-24 “inference requests from client applications”).
Wu and Giri are analogous to the claimed invention as they are all from the same field of endeavor of machine learning. Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing data of the claimed invention, to combine the inference response message and inference request message of Giri with the methodology of Wu. The motivation to do so is “to enable rapid execution of tasks (for example, rapid initialization of ML scoring container(s) 850, rapid execution of code 856 in ML scoring container(s), etc.) in response to deployment and/or execution requests” (Giri, Fig. 8 – 850, 856, Col. 23, Lines 47-51).
Regarding claim 9, Wu in view of Giri and further in view of Shu teaches the method of claim 1 (and thus the rejection of claim 1 is incorporated).
Regarding the limitation further comprising: prior to the requesting of the inference response message, creating the inference service according to a request from the client, Wu teaches the inference service (Fig. 1 – 102-106, ¶35). However, Wu fails to teach further comprising: prior to the requesting of the inference response message, creating the inference service according to a request from the client.
Giri teaches further comprising: prior to the requesting of the inference response message, creating a model according to a request from the client (Fig. 1 – Circle 2, 102, 104, 109, 112, Col. 6, Lines 44-49 “at circle (2) the computing device 104 may issue one or more requests (e.g., API calls)… that indicate the user's 109 desire to train a model(s) 112. The request may be of a type that identifies which type of model is to be created, e.g., CreateModel for creating a trained model 112,” Col. 7, Lines 23-24 “receive inference requests from client applications 140A and/or 140B at circle (7),” Fig. 1 – Circle 7 indicates that Circle 2 occurs prior to the requesting of the inference response message).
Regarding the limitation sequentially creating the inference container and the imitation learning container based on state information of the inference service, Wu teaches the inference container and the imitation learning container (Fig. 1 – 104 and 108) and the inference service (Fig. 1 – 102-106, ¶35 “The machine learning module… comprises a machine learning algorithm… The machine learning module 104 will calculate an output 106 based on the input 102”). However, Wu fails to teach sequentially creating the inference container and the imitation learning container based on state information of the inference service.
Maccanti teaches sequentially creating containers (Fig. 2 – 224, ¶45 “a storage system in which containers are stored. The repository 224 may storage many such containers, and each container may… be associated with a different machine learning model or endpoint,” Fig. 5 – 502-508, ¶63 “the system receives a request to create an endpoint to provide machine learning services. In embodiments, endpoint creation refers to the system enabling itself to receive requests to interact with a machine learning model… the endpoint is associated with a network address, to which requests to access the machine learning model are directed. In other embodiments, model servers associated with the endpoint are associated with network addresses,” ¶66 “the system generates and stores a container that comprises a model server associated with the requested endpoint and an extension,” wherein “storing” a container after its generation in a “repository” that stores “many such containers” implies sequentially creating multiple containers according to the process depicted in Figure 5) based on state information of an inference service (Fig. 5 – 508, ¶66 “generates and stores a container that comprises a model server associated with the requested endpoint and an extension,” Fig. 3 – 312-316, 340-342, ¶51 as explained above with respect to claim 3).
Regarding the limitation and performing a mirroring setting for sending the inference request message and the inference response message to the imitation learning container based on the state information, Wu teaches and performing a mirroring setting for sending input and output to the imitation learning container (Fig. 1 – 102-108, Fig. 2 – 204, ¶39 “the prediction explanation module 108 includes or is coupled to the second processor 204. For example, in one embodiment, the second processor 204 is configured to run the algorithm(s) of the prediction explanation module 108 by accessing at least the output 106 of the machine learning module 104, as well as the inputs 102,” wherein “is configured” implies a configuration, which encompasses a setting; one of ordinary skill in the art would recognize that a setting of the "the second processor” to “run the algorithm(s) of the prediction explanation module by accessing at least the output of the machine learning module, as well as the inputs” functions in substantially the same way as performing a mirroring setting). However, Wu fails to teach and performing a mirroring setting for sending the inference request message and the inference response message to the imitation learning container based on the state information.
Giri teaches the inference request message (Fig. 1 – 130, 134, 138, 140A-B, Col. 7, Lines 21-24) and the inference response message (Fig. 1 – 136, 140A-B, 160A-B, Col. 7, Lines 25-28). However, Giri fails to teach based on the state information.
Maccanti teaches interfacing based on the state information (Fig. 5 – 508, ¶66 “the extension refers to code that at least includes instructions for interfacing with the model server,” Fig. 3 – 312-316, 340-342, ¶51 “the extension 312 interfacing with the model server 314 to provide it with parameters 340, metadata 342, or other information to be used by the machine learning model 316,” as explained above with respect to claim 3).
Regarding the limitation and the state information includes first access information and input/output specifications of the inference container, Wu teaches first access information and input/output specifications of the inference container (Fig. 1 – 100-106, Fig. 2 – 208, ¶69, ¶51-53, ¶60, all as explained above with respect to claim 3). However, Wu fails to teach and the state information includes…
Maccanti teaches and the state information includes interfacing instructions (Fig. 5 – 508, ¶66, Fig. 3 – 312-316, 340-342, ¶51 as explained above with respect to claim 3).
Wu, Giri, and Maccanti are analogous to the claimed invention as they are all from the same field of endeavor of machine learning. Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing data of the claimed invention, to combine the inference response message, request from the client, and the inference request message of Giri and the sequential creation of containers and the state information of Maccanti with the methodology Wu. The motivation to do so is “to enable rapid execution of tasks (for example, rapid initialization of ML scoring container(s) 850, rapid execution of code 856 in ML scoring container(s), etc.) in response to deployment and/or execution requests” (Giri, Fig. 8 – 850, 856, Col. 23, Lines 47-51) and to be able to handle a surge in demand for machine learning services (Maccanti, ¶14-15 “Using a dedicated server may not work well if there is a surge in demand because the capacity of the dedicated server instance may be limited… this approach would typically require… administrative burden… To address these issues, the user is able to request that access to a machine learning model be provided using a serverless configuration”).
Claims 13, 14 and 19 are device claims that contain similar limitations to those of claims 3, 4 and 9, respectively. Therefore, claims 13, 14 and 19 are rejected under substantially the same rationale as claims 3, 4, and 9, respectively.
Claims 5 and 15 are rejected under U.S.C. 103 as being unpatentable over Wu in view of Giri and further in view of Shu, and further in view of Maccanti, and further in view of Sieckmann et al. (US 20210342569 A1, hereinafter Sieckmann).
Regarding claim 5, Wu in view of Giri and further in view of Shu, and further in view of Maccanti teaches the method of claim 3 (and thus the rejection of claim 3 is incorporated).
Regarding the limitation wherein the inference service descriptor further includes an interpretation field indicating provision of the interpretation information and second access information of the imitation learning container, Wu teaches provision of the interpretation information (Fig. 1 – 106, 110, ¶36 “to explain the reasoning for the output 106. The explanation output 110 is presented in a human understandable manner,” ¶42 “an explanation of the reasoning to arrive at the prediction decision 106 can be presented, such as on a user interface of a computing device”) and second access information of the imitation learning container (Fig. 1 – 100-106, Fig. 2 – 208, ¶69 “the network interface 208 can be configured to include or comprise a medium through which the machine learning module 104, the prediction explanation module 108… can communicate with each other. The communication network… may be a… wireless communication network… The devices of the system or apparatus 100 are potentially configured to connect to the communication network, in accordance with various wired and wireless communication protocols. Examples of such… wireless communication protocols may include… Transmission Control Protocol and Internet Protocol (TCP/IP),” one of ordinary skill in the art would recognize that an IP address, or second access information, is implicit in order for the “prediction explanation module,” or imitation learning container, to connect to a “communication network” in accordance with “Internet Protocol”). However, Wu fails to teach wherein the inference service descriptor further includes an interpretation field indicating provision of the interpretation information and second access information…
Maccanti teaches the inference service descriptor (¶66 “the system generates and stores a container that comprises a model server associated with the requested endpoint and an extension”). However, Maccanti fails to teach wherein the inference service descriptor further includes an interpretation field indicating provision of the interpretation information and second access information…
Sieckmann, in the same field of endeavor, teaches wherein metadata further includes an interpretation field indicating availability of a model (Fig. 10 – 1025, ¶90 “The container manager can find the suitable container image or produce a new one on the basis of the model metadata… The model manager outputs the desired model within a container,” ¶92 “Table 1 summarizes a few exemplary metadata and their purpose,” Row 13 of Table 1 on Page 24, ¶95 “Authors, Field (13), can make models available for free or for a fee”).
Regarding the limitation and the interpretation field includes a value indicating whether or not to activate an explanation field indicating the provision of the interpretation information, and the second access information is created according to an ingress setting and has an Internet Protocol (IP) address or a domain name of the imitation learning container for an Application Program Interface (API) setting, Wu teaches the provision of the interpretation information (Fig. 1 – 106, 110, ¶36, ¶42) and the second access information… has an Internet Protocol (IP) address or a domain name of the imitation learning container (Fig. 1 – 100-106, Fig. 2 – 208, ¶6 as explained above). However, Wu fails to teach and the interpretation field includes a value indicating whether or not to activate an explanation field indicating the provision of the interpretation information, and the second access information is created according to an ingress setting and has an Internet Protocol (IP) address or a domain name of the imitation learning container for an Application Program Interface (API) setting.
Giri teaches and an IP address is created according to an ingress setting and has an IP address for an Application Program Interface (API) setting (Fig. 9 – 900, Col. 27, Lines 40-56 “At least some public IP addresses may be allocated to… customers of the provider network 900… customer IP addresses may be assigned to resource instances by the customers, for example via an API provided by the service provider… customer IP addresses are allocated to customer accounts and can be remapped to other resource instances by the respective customers as necessary or desired… the customer controls that IP address until the customer chooses to release it,” wherein an ingress setting, when given its broadest reasonable interpretation, encompasses setting IP addresses to different resources according to customer necessities or desires, hence IP addresses that are “allocated to customer accounts” that can be “remapped to other resource instances by the respective customers” encompass is created according to an ingress setting; furthermore, “IP addresses… assigned to resource instances… via an API” encompass an Internet Protocol (IP) address for an Application Program Interface (API) setting). However, Giri fails to teach and the interpretation field includes a value indicating whether or not to activate an explanation field indicating the provision of the interpretation information…
Sieckmann teaches and the interpretation field includes a value indicating whether or not to activate an explanation field indicating availability of a model (Rows 13 and 14 in Table 1 on Page 24, ¶95 “Authors, Field (13), can make models available for free or for a fee. To this end, use can be made of different payment models, stored in Field (14), for example a one-time payment upon download or use-dependent payment in relation to the duration or frequency of use,” wherein it is implicit for the “Authors” field, or the interpretation field, to include a value, for example a Boolean value indicating whether or not a model is “available for free or for a fee,” indicating whether or not to activate an explanation field such as the “Payment model” field indicating when the model is available “for a fee”).
Wu, Giri, Maccanti, and Sieckmann are analogous to the claimed invention as they are all from the same field of endeavor of machine learning. Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing data of the claimed invention, to combine the inference service descriptor of Maccanti, the metadata fields of Sieckmann, and the IP address, ingress setting, and API setting of Giri with the methodology Wu. The motivation to do so is “to enable rapid execution of tasks (for example, rapid initialization of ML scoring container(s) 850, rapid execution of code 856 in ML scoring container(s), etc.) in response to deployment and/or execution requests” (Giri, Fig. 8 – 850, 856, Col. 23, Lines 47-51), to be able to handle a surge in demand for machine learning services (Maccanti, ¶14-15 “Using a dedicated server may not work well if there is a surge in demand because the capacity of the dedicated server instance may be limited… this approach would typically require… administrative burden… To address these issues, the user is able to request that access to a machine learning model be provided using a serverless configuration”), and to “facilitate making decisions in relation to a workflow in an efficient manner” (Sieckmann, ¶26).
Claim 15 is a device claim that contains similar limitations to those of claim 5. Therefore, claim 15 is rejected under substantially the same rationale as claim 5.
Claims 8 and 18 are rejected under U.S.C. 103 as being unpatentable over Wu in view of Giri and further in view of Shu, and further in view of Reed et al. (US 20050192992 A1, hereinafter Reed).
Regarding claim 8, Wu in view of Giri and further in view of Shu teaches the method of claim 1 (and thus the rejection of claim 1 is incorporated).
Regarding the limitation wherein the creating of the interpretation information and the providing of the created interpretation information to the client further includes continuously training the imitation learning container until the critical condition is satisfied to output the imitation inference result, creating interpretation information of the imitation inference result that satisfies the critical condition, and discarding the imitation inference result that does not satisfy the critical condition, Wu teaches wherein the creating of the interpretation information (Fig. 1 – 102-110, ¶39 “the prediction explanation module 108 by accessing at least the output 106 of the machine learning module 104, as well as the inputs 102 and generate the explanation output 110”) and the providing of the created interpretation information to the client (Fig. 3 – 304-308, ¶42 “an explanation of the reasoning to arrive at the prediction decision 106 can be presented, such as on a user interface of a computing device”) further includes… training the imitation learning container… to output the imitation inference result (Fig. 1 – 102-108, Fig. 2 – 202-204, ¶63 “the casual relationships generated by the second processor 204 are based on imitative training of another machine learning algorithm with access to the machine learning algorithm in the first processor 202. This means that the second processor 204 has learned to generate the same or very similar predictions compared to the machine learning algorithm 104 run by the first processor 202, given the same input 102,” ¶52 “The input images are segmented into groups of pixels, also referred to herein as “visual concepts.” The groups of pixels are input into a learned machine/deep model, such as the model implemented by the prediction explanation module 108. The top “n” visual concepts can be selected based on the output of the machine/deep model by ranking the score each visual concept corresponds to”), creating interpretation information of the imitation inference result (Fig. 1 – 108-110, ¶52 “This ranking is presented by the explanation output 110,” ¶36 “The explanation output 110 is presented in a human understandable manner”). However, Wu fails to teach wherein the creating of the interpretation information and the providing of the created interpretation information to the client further includes continuously training the imitation learning container until the critical condition is satisfied to output the imitation inference result, creating interpretation information of the imitation inference result that satisfies the critical condition, and discarding the imitation inference result that does not satisfy the critical condition.
Shu teaches continuously updating a second model until the critical condition is satisfied to output a vector that satisfies the critical condition (¶74 “it is necessary to make the feature vector output by the last hidden layer of the second model and the feature vector output by the last hidden layer of the first model most similar… if the similarity coefficient does not meet the similarity parameter target value, continuing to adjust the second model according to the current similarity parameter until the calculated similarity parameter meets the similarity parameter target value, and stopping adjusting”). However, Shu fails to teach and discarding the imitation inference result that does not satisfy the critical condition.
Reed, in the same field of endeavor, teaches and discarding data that does not satisfy a condition (Fig. 5 – 530-545, ¶44 “At 530, it can be determined whether there is the intent to do something or merely an idle intent. Such determination can be based on a comparison with predefined threshold. If at 530 it is determined that no further action is desired, the data can be discarded”).
Wu, Shu, and Reed are analogous to the claimed invention as they are all from the same field of endeavor of machine learning. Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing data of the claimed invention, to combine the satisfying of the critical condition of Shu and the discarding of the result that does not satisfy the condition of Reed with the methodology of Wu. The motivation to do so is to improve “convenience and accuracy of model processing” (Shu, ¶44) and to implement “methods that determine intent for received data… to yield very high accuracy” (Reed, Abstract).
Claim 18 is a device claim that contains similar limitations to those of claim 8. Therefore, claim 18 is rejected under substantially the same rationale as claim 8.
Claim 10 is rejected under U.S.C. 103 as being unpatentable over Wu in view of Giri and further in view of Shu, and further in view of Maccanti, and further in view of Mueller et al. (US 20210326717 A1, hereinafter Mueller).
Regarding claim 10, Wu in view of Giri and further in view of Shu, and further in view of Maccanti teaches the method of claim 9 (and thus the rejection of claim 9 is incorporated).
Regarding the limitation wherein the creating of the inference container includes creating a sub-resource including an ingress layer, a service layer, and an inference container based on the inference service, Wu teaches the inference container (Fig. 1 – 104) and the inference container based on the inference service (Fig. 1 – 102-106, ¶35 “The machine learning module 104… comprises a machine learning algorithm… The machine learning module 104 will calculate an output 106 based on the input 102”). However, Wu fails to teach wherein the creating of the inference container includes creating a sub-resource including an ingress layer, a service layer, and an inference container based on the inference service.
Mueller, in the same field of endeavor, teaches wherein the creating of a container includes creating a sub-resource including an ingress layer (Fig. 10 – 1050, ¶127 “another system (e.g., a routing system, not shown) can obtain the execution request, identify the ML scoring container(s) 1050 corresponding to the identified endpoint, and route the input to the identified ML scoring container(s),” given its broadest reasonable interpretation, an ingress layer controls entry points for inbound traffic; one of ordinary skill in the art would recognize that a “routing system” that routes “input” to “identified ML scoring container(s)” functions in substantially the same way as an ingress layer), a service layer (Fig. 10 – 140, 1002, 1050, ¶126 “The model hosting system 140 can map the network address(es) to the identified endpoint… a user device 1002 can refer to trained machine learning model(s) stored in the ML scoring container(s) 1050 using the endpoint,” given its broadest reasonable interpretation, a service layer maps network addresses to their appropriate services; one of ordinary skill in the art would recognize that a “model hosting system” functions in substantially the same way as a service system)…
Wu and Mueller are analogous to the claimed invention as they are all from the same field of endeavor of machine learning. Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing data of the claimed invention, to combine the sub-resource including an ingress layer and a service layer of Mueller with the methodology of Wu. The motivation to do so is to allow “users to easily train high-quality custom machine learning (ML) models and/or pipelines for without necessarily needing to write code or have significant knowledge of ML concepts or techniques” (Mueller, ¶18).
Response to Amendment
In review of Applicant’s amendments, filed April 6, 2026, the objections to the specification and drawings made in the previous office action have been withdrawn.
The rejections of claims 5, 6, 11, 15, 16, and 20 under 35 U.S.C. 112(b) set forth in the previous office action are withdrawn in view of the amendments to the claims.
Claim 20 is no longer being interpreted under 35 U.S.C. 112(f) in view of applicant’s amendments to the claim.
Response to Arguments
Applicant’s arguments, see page 14 section V filed April 6, 2026, regarding the rejections from the previous office action made under 35 U.S.C. 103 have been fully considered but are moot as they do not apply to the reference Shu being used in the current rejections of claims 1, 11, and 20 and their associated dependent claims 2-10 and 12-19, respectively, to teach the amended claim limitations directed to “the creating of the interpretation information…,” “the imitation inference result includes an inference derived…,” and “the critical condition is set to a condition…”
Specifically, Shu teaches “the critical condition” based on “a similarity of a reference value or greater” (¶74).
Additionally, the existing reference Wu teaches “the inference result includes an inference derived from object information provided by the client” (¶61, ¶65).
With the existing reference Wu and the addition of the reference Shu teaching the subject matter introduced in the amendments, the rejections under 35 U.S.C. 103 stand.
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 WILLIAM M LEE whose telephone number is (571)272-4761. The examiner can normally be reached Mon-Fri. 8am-5pm.
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.
/WILLIAM MICHAEL LEE/
Examiner, Art Unit 2145
/CESAR B PAULA/ Supervisory Patent Examiner, Art Unit 2145
Prosecution Insights