DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
This action is in response to the amend lent filed on 09/17/2025. Claims 1-20 are pending in the case. This action is Final.
Applicant Response
In Applicant’s response dated 09/17/2025, Applicant amended Claims 1, 2, 8, 11, 12, 16, and 17 and argued against all objections and rejections previously set forth in the Office Action dated 06/18/2025.
Claim Rejections - 35 USC § 101
4. 35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
5. Claims 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Regarding Claim 1,
Claim 1 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 1 is directed to a computer implemented method, which is directed to a process, one of the statutory categories.
Step 2A Prong One Analysis: The limitations:
“Identify an occurrence of an inference model update event for an inference model of the inference models wherein the inference model update event comprises the inference model failing to meet a threshold for accurately identifying features in a first portion of the collected unlabeled data, the inference model is a neural network, and the collected data is unlabeled data; (This step involve measuring model performance, an evaluation of data using mathematical model (i.e., evaluation/math.).);
in response to the inference model update event: generate an inference model update package for the inference model, the inference model update package comprising a result of a calculation performed on a first portion of the unlabeled collected data usable to partially adapt the inference model to detect a feature in the first portion of the unlabeled collected data, an unlabeled data update and an identifier wherein the result is less than 0.1% of the size of the first portion of the unlabeled collected data and all sensitive data in the first portion of the unlabeled collected data is not included in the result, wherein the sensitive data includes at least one of private data and confidential data; ; (This step involve mathematical operation on data (i.e., Mathematical concept, data compression.).);
obtain, from the entity, a hybrid data adapted inference model that is based on: the inference model, the inference model update package, and labeled data used to train the inference mode and stored in the entity (This step involve mathematical model manipulation operation on data (i.e., Mathematical concept.); and
obtain an inference, using the hybrid data adapted inference model and a second portion of the unlabeled collected data, that indicates a second feature is present in the second portion of the collected data. (This step involve mathematical operation on data (i.e., Mathematical concept.).);”
As drafted, under their broadest reasonable interpretations, cover mathematical operations, mathematical concepts and data transformation steps. The claim recites a mathematical calculation which fall within the mathematical concepts grouping of abstract ideas. Accordingly, the claims recite an abstract idea.
Step 2A Prong Two Analysis: Does the claim recite additional elements? Do those additional elements, individually and in combination, integrate the judicial exception into a practical application?
The claim recites additional elements:
a plurality of cameras programmed to collect unlabeled data, wherein the plurality of cameras are connected to a car ( The car cameras are generic computer/field of use of limitation, not meaningful technological improvement. Putting a car with camera is just applying it to a known environment.)
the inference model is a neural network, and the collected data is unlabeled data; the result is less than 0.1% a size of the first portion of the collected data, and all sensitive data in the first portion of the collected data is not included in the result, wherein the sensitive data includes at least one of private data and confidential data;
provide the inference model update package to an entity that generated the inference model; obtain, from the entity, a hybrid data adapted inference model that is based on: the inference model, the inference model update package, and labeled data used to train the inference model and stored in the entity, and wherein, the hybrid data adapted inference model is a neural network; and obtain an inference by: using a second portion of the collected data as an input to the hybrid data adapted inference model; and
receiving the inference as an output, based on the input, from the hybrid data adapted inference model, wherein the inference specifies a set of features of the second portion of the collected data.
These steps involve analyzing data, performing calculations and generating a model update and can be performed on a generic computing device. The processor, storage model are general purpose computing elements. In addition, the requirement that the calculation is “less than 1% size” and “excluding sensitive data” are a desired result and does not constitute a technical process or solution. The core invention is about updating a neural network using local data , compressed updates and remote retracing and having a car with camera or compressing updates (data manipulation rule)) are nothing more than limiting the abstract idea to a practical field of use. Therefore, the additional limitations fail to integrate the abstract idea into a practical application. The limitations, recited at a high level of generality, only amount to “apply it” using a generic computer component (MPEP 2106.05(f)).
Step 2B Analysis: Do the additional elements, considered individually and in combination, amount to significantly more than the judicial exception?
No, the claim recites additional element “a plurality of cameras programmed to collect unlabeled data, wherein the plurality of cameras are connected to a car ( The car cameras are generic computer/field of use of limitation, not meaningful technological improvement. Putting a car with camera is just applying it to a known environment.), “provide the inference model update package to an entity that generated the inference model.” These additional limitations fail to integrate the abstract idea into a practical application. The limitations, recited at a high level of generality, only amount to “apply it” using a generic computer component (MPEP 2106.05(f)).
The dependent claims respectively recite a judicial exception in limitations of: “identifying the occurrence of the inference model update event: obtain a second inference, using a labeled data adapted inference model of the inference models and the second portion of the unlabeled collected data, that indicates that the set of features are second feature is second feature is not present in the second portion of the unlabeled collected data (claims 2, 12 and 17), “wherein the labeled data adapted inference model is not based on the unlabeled data” (claims 3, 13 and 18), “wherein the unlabeled data comprises features, for which the inference models are adapted to identify, which are not identified in the unlabeled data” (claims 4, 14 and 19), “wherein the labeled data comprises features, for which the inference models are adapted to identify, which are identified in the labeled data” (claims 5, 15 and 20), “wherein the portion of the collected data consists of unlabeled data”, “wherein the result comprises a gradient and a target loss for a neural network”, “wherein the hybrid data adapted inference model is obtained prior to the portion of the collected data being provided to the entity” (claim 10), “wherein the inference model update package further comprises the inference model, (claim 9) “wherein the inference model update event is a change in a natural environment in which the information handling system is disposed.”(Claim 10).
These additional limitations (in claims 2-9, 11-15 and 17-20) also constitute concepts performed in the human mind which fall within the “Mental Processes” groupings of abstract ideas.
This judicial exception is not integrated into a practical application. Additional elements “non-transitory computer readable medium comprising: computer program code” (in claims 2-11-15 and 17-20), all amount to no more than adding insignificant extra-solution activity/specifications related to data gathering, data input, or data transmittal. These additional elements do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea. The dependent claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, the additional elements of non-transitory computer readable medium comprising: computer program code are again insignificant extra-solution activity steps that cannot provide an inventive concept. All of these additional elements as generically claimed are considered well-understood, routine, and conventional.
Therefore, these limitations, taken alone or in combination, do not integrate the abstract idea into a practical application or recite significantly more that the abstract idea. Thus, all of the dependent claims are also not patent eligible.
Examiner Comments
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1-6 and 8-20 are rejected under 35 U.S.C. 103 as being unpatentable over Anderson (Pub. No.: US 20190220028 A1, Pub. Date: 2021-09-30) in view of CROSBY (Pub. No.: US 20220138509 A1, Pub. Date: 2022-05-05) in further view of McMahan (Pub. No.: US 20190340534 A1, Pub. Date 2019-11-07) in further view of Malekijoo (NPL: Title: FEDZIP: a compression framework for communication-efficient federated learning: Pub; Date 2021-02-02)
Regarding independent Claim 1,
Anderson teaches an information handling system, comprising:
a plurality of cameras programmed to collect unlabeled data, wherein the plurality of cameras are connected to a car (see Anderson: Fig.1, [0027], “In addition to position sensors on vehicle 102's controls, in embodiments, apparatus 104 is in communication with and receives data (i.e. unlabeled data) from various data sources of sensors 110 ( i.e. for example camera). Sensors 110 may include video, audio, LIDAR, radar, ultrasonic, and similar types of detectors or sensors for the external environment of vehicle 102, as well as internal sensors such as speed, acceleration, … and equipment status such as whether the brakes are applied, the throttle is applied, turn signal actuation, anti-lock brake system status, rain sensors, wiper actuation, and/or any other similar data that may be used by vehicle 102 to navigate around vehicle 102's environment, and are also useable to determine the presence of a problematic or trouble spot 150. It will be understood that the types of data will depend upon the nature of the data source.”)
storage for storing (see Anderson: Fig.2, [0040], “the server 202 may include a database 210, which can act as a storage repository for data from the data sources):
the collected unlabeled data (see Anderson: Fig.1, [0022], “transmits data collected from the one or more sensors 110 and the data on the driver actions to a remote server 120); and
inference models (see Anderson: Fig.2, [0036], an inference modeler 204);
a processor (see Anderson: Fig.2: FIG. 7, an example computing platform with a processor), programmed to:
identify an occurrence of an inference model update event for an inference model of the inference models (see Anderson: Fig.4: [0059], “In operation 406, the inference model is tested, using the received data and in comparison, with human driver action data also collected, to identify any problem locations. The identified problem location may be iteratively tested using subsets of available data, including subsets of variable data, such as weather conditions, traffic conditions, train or other vehicle schedules, etc., to attempt to isolate and identify, in operation 408, one or more additional or different data categories to collect. The additional or different data categories may allow improving the inference model to correctly and reliably handle the identified problem location.”, i.e., additional or different data categories is considered as the inference model update event.), wherein:
the inference model update event comprises the inference model failing to meet a threshold for accurately identifying features in a first portion of the unlabeled collected data (see Anderson: Fig.4, [0059], “he identified problem location may be iteratively tested using subsets of available data, including subsets of variable data, such as weather conditions, traffic conditions, train or other vehicle schedules, etc., to attempt to isolate and identify, in operation 408, one or more additional or different data categories to collect.”.. (see Anderson: Fig.4, [0057], “Starting in operation 402, data is received relating to a plurality of categories for a driving environment.”, i.e., the data received is unlabeled data and is stored in the CA/AD vehicle navigation system),
the inference model is a neural network (see Anderson: Fig.4, [0062], “apparatus 104, and in particular, system 200 (including server 202, inference modeler 204, and simulator(s) 206), may include one or more trained neural networks in performing its determinations and/or assessment.”) and
in response to the inference model update event (see Anderson: Fig.4, [0059], “e or more additional or different data categories to collect. The additional or different data categories may allow improving the inference model to correctly and reliably handle the identified problem location”):
generate an inference model update package for the inference model (see Anderson: Fig.4, [0061], “a second or subsequent inference model is generated using the data from the identified category/categories, and possibly using the existing inference model and/or earlier data set. This second or subsequent inference model may then be passed through method 400 again in an iterative fashion, to potentially generate further, more refined or more accurate inference models.”), […]
provide the inference model update package to an entity that generated the inference model (see Anderson: Fig.4, [0058], “the inference model may be generated by training a NN, such as a DNN or CNN, with the received data, and then optimizing the trained NN to obtain a targeted inference model designed to analyze driving related data useful for CA/AD vehicle navigation.”)
obtain, from the entity, a hybrid data adapted inference model (see Anderson: Fig.4, [0058], “operation 404, the received data is used to generate a first version of an inference model.”, i.e., the first version inference model is the hybrid data adapted inference model), that is based on:
the inference model (see Anderson: Fig.4, [0055], “If the inference model is determined to be substantially or entirely free of trouble spots (at least those capable of being identified and tested with available data sources and data sets), the inference model may be pushed out to one or more vehicles, such as vehicle 102, that are in communication with an implementing system, such as system 200.”),
the inference model update package (see Anderson: Fig.4, [0059], “The additional or different data categories may allow improving the inference model to correctly and reliably handle the identified problem location.”, i.e., additional or different data categories is considered as the inference model update event.), and
labeled data used to train the inference mode and stored in the entity, and wherein the hybrid data adapted inference model is a neural network (see Anderson: Fig.4, [0057], “Starting in operation 402, data is received relating to a plurality of categories for a driving environment.”, i.e., the data received is labeled data and is stored in the CA/AD vehicle navigation system); and
obtain an inference (see Anderson: Fig.4, [0058], “operation 404, the received data is used to generate a first version of an inference model.”, i.e., the first version inference model is the hybrid data adapted inference model), by:
using a second portion of the unlabeled collected data as an input to the hybrid data adapted inference model (see Anderson: Fig.4, [0061], “In operation 412, a second or subsequent inference model is generated using the data from the identified category/categories, and possibly using the existing inference model and/or earlier data set.”); and
receiving the inference as an output, based on the input, from the hybrid data adapted inference model, wherein the inference specifies a set of features of the second portion of the collected unlabeled data (see Anderson: Fig.1, [0022], “receive from the remote server 120 a subsequent refined or improved inference model, the second inference model generated from the first inference model updated with the data from the one or more sensors 110 and the data on the driver actions. Remote server 120 may be implemented as a part of system 200, and in particular a type of server 202, which will be described below with respect to FIG. 2.”)
As shown above, Anderson teaches or discloses an improvement to autonomous driving systems through the use of artificial intelligence (AI), using neural networks to generate an inference model that allows the NN to solve problems against data sets that may differ somewhat from the training sets, such as by inferring an appropriate response from roughly similar training data.
Anderson does not teach the method wherein:
the inference model update package comprising a result of a calculation performed on a first portion of the unlabeled collected data usable to partially adapt the inference model to detect a feature in the first portion of the unlabeled collected data, an unlabeled data update, and an identifier wherein the result is less than 0.1% a size of the first portion of the collected data; and
all sensitive data in the first portion of the collected unlabeled data is not included in the result, wherein the sensitive data includes at least one of private data and confidential data;
However, CROSBY teach the method wherein the inference model update package comprise a result of a calculation performed on a portion of the collected data unlabeled usable to partially adapt the inference model to detect a feature in the first portion of the collected unlabeled data an unlabeled data update, and an identifier (see CROSBY: Fig.3, [0037], “Unsupervised feature extraction module 107a includes two modules that perform unsupervised feature extraction processes: cleaning and grading module 300 that outputs cleaned data 301 (inference model update package), and self-supervised pretext task learning 302 that generates learned intermediate features 303. The outputs from the unsupervised feature extraction module, learned intermediate features 303 and clean unlabeled data 301 then are used as inputs supervised learning module 112a.”, i.e. the cleaning/grading perform calculation on the collected data.) … see also [0038], Cleaning and grading module 300 analyzes the data in unlabeled dataset 201 to determine if the data meets certain minimum criteria to be usefully analyzed by the self-supervised pretext task-learning module 302. In particular, a grading analysis is performed on unlabeled dataset 201 that filters the unlabeled data and generates as an output cleaned data 301”)
Because both Anderson and CROSBY are in the same/similar field of endeavor of improvement of machine learning model, accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the invention to modify the teaching of Anderson to include an updated data package that comprise the result of a calculation performed on a portion of the collected data usable to partially adapt the inference model to detect a feature in the portion of the collected data package as taught by CROSBY. After modification of Anderson, the computer-assisted and autonomous driving system that process data received relating to a plurality of categories for a driving environment can also analyses the collected data using the cleaning and grading module to determine feature in the portion of the collected data as taught by CROSBY. One would have been motivated to make such a combination in order to provide users with efficient, autonomous and continually self-improving learning system that is fully scalable, low-cost, time-efficient, and highly effective in its performance. (see CROSBY [0020])
McMahan teaches the system wherein:
all sensitive data in the first portion of the collected unlabeled data is not included in the result, wherein the sensitive data includes at least one of private data and confidential data (see McMahan: Fig.1, [0059], “Client devices 102 can be configured to provide the local updates to server 104. As indicated above, training data 108 may be privacy sensitive. In this manner, the local updates can be performed and provided to server 104 without compromising the privacy of training data 108. For instance, in such implementations, training data 108 is not provided to server 104.” …. [0021], “Federated learning offers several distinct advantages compared to performing learning at a centralized server. For example, information of the model update is less sensitive that the data itself. Thus, user data that is privacy sensitive remains at the user's computing device and is not uploaded to the server. Instead, only the less sensitive model update is transmitted.”)
Because Anderson, CROSBY and McMahan are in the same/similar field of endeavor updating machine learning model, accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the invention to modify the teaching of Anderson to include an updated data package that comprise the result of a calculation performed on a portion of the collected data usable to partially adapt the inference model to detect a feature in the portion of the collected data package and a package that preserve sensitive data taught by McMahan. One would have been motivated to make such a combination in order to increase communication efficiency by providing users with a system that both protects sensitive data and reduce communication payload. (see McMahan [0009])
Anderson, CROSBY and McMahan does not explicitly teach the system wherein the calculation result is less than 0.1% a size of the first portion of the collected unlabeled data.
However, Malekijoo teaches wherein the calculation result is less than 0.1% a size of the first portion of the collected data (see Abstract: Page1, Line 12-15, “FedZip implements Top-z sparsification, uses quantization with clustering, and implements compression with three different encoding methods. FedZip outperforms state-of-the-art compression frameworks and reaches compression rates up to 1085, and preserves up to 99% of bandwidth and 99% of energy for clients during communication.”, i.e. a 1085x compression corresponds to 0.092% (1/1085), which is below the 0.1% size constraint).
Because Anderson, CROSBY, McMahan and Malekijoo are in the same/similar field of endeavor of training a machine learning model, accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the invention to modify the teaching of Anderson to include an updated data package calculated result less than 0.1% a size of the first portion of the collected data as taught by Malekijoo. One would have been motivated to make such a combination in order to provide provides improved techniques to significantly decreases the size of updates while transferring weights from the deep learning model between clients and their servers. (see Abstract)
Regarding Claim 2,
Anderson, CROSBY, McMahan and Malekijoo teach all the limitations of claim 1. Anderson further teaches the processor is further programmed to:
prior to identifying the occurrence of the inference model update event (see Anderson: Fig.4, [0060], “following identification of the one or more additional or different data categories, data for the identified category/categories is collected, such as by dispatching other vehicles and/or configuring any stationary sources or road controllers to obtain data in the identified category/categories.”), obtain a second inference, using a labeled data adapted inference model of the inference models and the second portion of the collected unlabeled data, that indicates that the set of features are is not present in the second portion of the collected unlabeled data see Anderson: Fig.4, [0061], “in operation 412, a second or subsequent inference model is generated using the data from the identified category/categories, and possibly using the existing inference model and/or earlier data set. This second or subsequent inference model may then be passed through method 400 again in an iterative fashion, to potentially generate further, more refined or more accurate inference models.”)
Regarding Claim 3,
Anderson, CROSBY, McMahan and Malekijoo teaches all the limitations of claim 2. CROSBY further teaches the system wherein the labeled data adapted inference model is not based on the unlabeled data (see CROSBY: Fig.2, [0036], “The labeled curated dataset may be further partitioned at data splitter step 204, with a portion of curated dataset 203 being made accessible to supervised learning module 112a via processed supervisor bucket 110, and the remainder of the subset being reserved for use by validation module 114a. The entire unlabeled dataset 201 also is processed by unsupervised feature extraction module 107a.”)
One would have been motivated to combine Anderson and CROSBY, before the effective filing date of the invention because it provides the benefit where self-improving learning system that is fully scalable, low-cost, time-efficient, and highly effective in its performance. (See CROSBY [0020])
Regarding Claim 4,
Anderson, CROSBY, McMahan and Malekijoo teaches all the limitations of claim 3. CROSBY further teaches the system wherein the unlabeled data comprises features, for which the inference models are adapted to identify, which are not identified in the unlabeled data (see CROSBY: Fig.2, [0038], “Cleaning and grading module 300 analyzes the data in unlabeled dataset 201 to determine if the data meets certain minimum criteria to be usefully analyzed by the self-supervised pretext task-learning module 302. In particular, a grading analysis is performed on unlabeled dataset 201 that filters the unlabeled data and generates as an output cleaned data 301.”)
One would have been motivated to combine Anderson and CROSBY, before the effective filing date of the invention because it provides the benefit where self-improving learning system that is fully scalable, low-cost, time-efficient, and highly effective in its performance. (See CROSBY [0020])
Regarding Claim 5,
Anderson, CROSBY, McMahan and Malekijoo teaches all the limitations of claim 4. CROSBY further teaches the system wherein the labeled data comprises features, for which the inference models are labeled adapted to identify, which are identified in the labeled data (see CROSBY: Fig.2, [0038], “. The labeled curated dataset may be further partitioned at data splitter step 204, with a portion of curated dataset 203 being made accessible to supervised learning module 112a via processed supervisor bucket 110, and the remainder of the subset being reserved for use by validation module 114a. The entire unlabeled dataset 201 also is processed by unsupervised feature extraction module 107a”)
See Claim 1 above, for motivation to combine Anderson and CROSBY.
Regarding Claim 6,
Anderson, CROSBY, McMahan and Malekijoo teaches all the limitations of claim 3. CROSBY further teaches the system wherein the portion of the collected unlabeled data consists of unlabeled data (see CROSBY: Fig.2, [0036], “multiple datasets 200 are stored on mass storage 105 described in FIG. 1. A particular dataset selected from mass data storage 105 initially consists of unlabeled data 201”)
See Claim 1 above, for motivation to combine Anderson and CROSBY.
Regarding Claim 8,
Anderson, CROSBY, McMahan and Malekijoo teaches all the limitations of claim 1. CROSBY further teaches the system wherein the hybrid data adapted inference model is obtained prior to the portion of the collected unlabeled data being provided to the entity (see Anderson: Fig.4, [0061], “In operation 412, a second or subsequent inference model is generated using the data from the identified category/categories, and possibly using the existing inference model and/or earlier data set. This second or subsequent inference model may then be passed through method 400 again in an iterative fashion, to potentially generate further, more refined or more accurate inference models.”)
One would have been motivated to combine Anderson and CROSBY, before the effective filing date of the invention because it provides the benefit where self-improving learning system that is fully scalable, low-cost, time-efficient, and highly effective in its performance.
Regarding Claim 9,
Anderson, CROSBY, McMahan and Malekijoo teaches all the limitations of claim 1. CROSBY further teaches the system wherein the inference model update package further comprises the inference model (see Anderson: Fig.4, [0060], “In operation 410, following identification of the one or more additional or different data categories, data for the identified category/categories is collected, such as by dispatching other vehicles and/or configuring any stationary sources or road controllers to obtain data in the identified category/categories.”)
Regarding Claim 10,
Anderson, CROSBY, McMahan and Malekijoo teaches all the limitations of claim 1. CROSBY further teaches the system wherein the inference model update event is a change in a natural environment in which the information handling system is dispose (see Anderson: Fig.4, [0057], “Starting in operation 402, data is received relating to a plurality of categories for a driving environment. Categories may include traffic conditions, signage, vehicles, crossings, weather conditions, road construction, and any other category of information that may be used by a CA/AD vehicle to analyze and determine a navigation solution.”)
One would have been motivated to combine Anderson and CROSBY, before the effective filing date of the invention because it provides the benefit where self-improving learning system that is fully scalable, low-cost, time-efficient, and highly effective in its performance. (See CROSBY [0020])
Regarding independent Claim 11,
Claim 11 is directed to a method claim and has the same/similar claim limitation as claim 1 and is rejected under the same rationale.
Regarding Claim 12,
Claim 12 is directed to a method claim and has the same/similar claim limitation as claim 2 and is rejected under the same rationale.
Regarding Claim 13,
Claim 13 is directed to a method claim and has the same/similar claim limitation as claim 13 and is rejected under the same rationale.
Regarding Claim 14,
Claim 14 is directed to a method claim and has the same/similar claim limitation as claim 4 and is rejected under the same rationale.
Regarding Claim 15,
Claim 15 is directed to a method claim and has the same/similar claim limitation as claim 5 and is rejected under the same rationale.
Regarding Claim independent 16,
Claim 16 is directed to a method claim and has the same/similar claim limitation as claim 1 and is rejected under the same rationale.
Regarding Claim 17,
Claim 17 is directed to non-transitory computer readable medium claim and has the same/similar claim limitation as claim 2 and is rejected under the same rationale.
Regarding Claim 18,
Claim 18 is directed to non-transitory computer readable medium claim and has the same/similar claim limitation as claim 3 and is rejected under the same rationale.
Regarding Claim 19,
Claim 19 is directed to non-transitory computer readable medium claim and has the same/similar claim limitation as claim 4 and is rejected under the same rationale.
Regarding Claim 20,
Claim 20 is directed to non-transitory computer readable medium claim and has the same/similar claim limitation as claim 5 and is rejected under the same rationale.
Claim 7 is rejected under 35 U.S.C. 103 as being unpatentable over Anderson in view of CROSBY, McMahan and Malekijoo as applied to claims 1-6 and 8-20 as shown above and in further view of Karlinsky (Pub. No.: US 20220188639 A1, Pub. Date: 2022-06-16)
Regarding Claim 7,
As shown above, Anderson, CROSBY, McMahan and Malekijoo teach all the limitations of Claim 1. Anderson and CROSBY does not teach the system wherein the result comprises a gradient and a target loss for a neural network.
However, Karlinsky teaches the system wherein the result comprises a gradient and a target loss for a neural network (see Karlinsky: Fig.5, [0037], “Gradient generating subnetwork 106 may be an additional component to common subnetwork 104 and configured to compute a gradient. Training module 110 may determine gradient generating loss 310 of gradient generating subnetwork 106 based on input data 302 (e.g., unlabeled data). Training module 110 may use gradient generating loss 310 to compute the gradient of gradient generating subnetwork 106. Training module 110 may generate, from unlabeled data samples, gradients that will help reduce main task loss 306 in common subnetwork 104.”)
Tt would have been obvious to a person of ordinary skill in the art, before the effective filing date of the invention to modify the teaching of Anderson to include system wherein the result comprises a gradient and a target loss for a neural network as taught by CROSBY. One would have been motivated to make such a combination in order to provide users with efficient, accurate machine learning model by lowering error resulting better the model's predictions.
Response to Amendment
Claim Rejections - 35 U.S.C. § 101,
Regarding the 35 U.S.C. 101 rejection for being directed non-statutory subject matter has been maintained based on the new amendment.
Response to Arguments
Applicant's prior art arguments with respect to the currently amended independent claims and the dependent claims have been fully considered but they are moot in view of the new grounds of rejection presented above. Applicant is respectfully referred to the complete rejections presented above and the newly cited portions of the references previously relied upon. Examiner further notes that Applicant’s arguments are mere allegations that the cited art does not teach the limitations of the independent claim as amended and do not explicitly show any deficiencies with the previously cited art of the record in relationship with the newly recited limitations.
Thus, Examiner respectfully reasserts that the combination of Anderson in view of CROSBY, McMahan and Malekijoo sufficiently teaches all the limitations recited in the independent claims, as amended, and therefore the claims and 10-12 are still rejected under 35 U.S.C. 103 as being unpatentable over the prior arts shown above.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
PGPUB
NUMBER:
INVENTOR-INFORMATION:
TITLE / DESCRIPTION
US 20230186637 A1
Singh; Gurjeet
Title: SYSTEMS AND METHODS FOR DETECTING DEEP NEURAL NETWORK INFERENCE QUALITY USING IMAGE/DATA MANIPULATION WITHOUT GROUND TRUTH INFORMATION
Description: This disclosure generally relates to artificial intelligence using deep neural networks, and more particularly relates to systems and methods for detecting deep neural network (DNN) inference quality using image/data manipulation without ground truth information.
US 20190258878 A1
Dunne; Aubrey
Title: OBJECT DETECTION AND DETECTION CONFIDENCE SUITABLE FOR AUTONOMOUS DRIVING
Description: present disclosure relate to object detection and detection confidence suitable for autonomous driving. In contrast to conventional approaches that determine a confidence value for an aggregated detection that may not be directly interpreted as a confidence or probability measure or may only correspond to a small portion of an image.
US 20230221942 A1
Zhang; Wende
Title: FEDERATED LEARNING FOR CONNECTED CAMERA APPLICATIONS IN VEHICLES
Description: The technical field generally relates to vehicles and more particularly relates to automated incident detection for vehicles based on captured images using federated learning.
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 ZELALEM W SHALU whose telephone number is (571)272-3003. The examiner can normally be reached M- F 0800am- 0500pm.
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.
Zelalem Shalu
Examiner
Art Unit 2145
/Zelalem Shalu/Examiner, Art Unit 2145
/CESAR B PAULA/Supervisory Patent Examiner, Art Unit 2145