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 office action is in response to submission of application on 12/20/2023.
Claims 1-20 are presented for examination.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-20 are rejected under 35 U.S.C. § 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1: Is the claim to a process, machine, manufacture, or composition of matter?
Claims 1-8 are directed to a method (i.e., a process), claims 9-18 are directed to a system (i.e., a machine/apparatus), and claims 19-20 are directed to a non-transitory computer readable medium(i.e., a product/article of manufacture); therefore, all pending claims are directed to one of the four categories of invention.
Step 2A: Prong 1: Does the claim recite an abstract idea, law of nature, or natural phenomenon?
Claim 1 recites limitations of:
generating updated training data for the machine learning model based, at least in part, on the embedding of the item of data and the user action with respect to the item of data – mental process (observation, evaluation, judgement, opinion) as a human mind can generate updated training data for a machine learning model.
which are abstract idea, something that can be accomplished by a human mind, or with the aid of pen and paper.
Step 2A: Prong 2: Does the claim recite additional elements that integrate the judicial exception into a practical application?
Claim 1 recites the additional elements of
A method for re-training a machine learning model in real-time – machine learning models recited at a high level are construed as generic models used to implement the abstract idea. See MPEP 2106.05(f)(1).
generating an embedding of an item of data associated with a user of a software application – generating an embedding is merely preparing data so that it can be input to a computer. As such, it is insignificant, extra-solution activity. See MPEP 2106.05(g).
storing the embedding of the item of data on a data store prior to a trigger event comprising a user action with respect to the item of data – storing data is mere data gathering which is insignificant, extra-solution activity. See MPEP 2106.05(g).
obtaining data indicative of the trigger event during an online session for the user of the software application – obtaining data is mere data gathering which is insignificant, extra-solution activity. See MPEP 2106.05(g).
retrieving the embedding of the item of data from the data store in response to obtaining the data indicative of the trigger event during the online session – retrieving data is mere data gathering which is insignificant, extra-solution activity. See MPEP 2106.05(g).
providing the updated training data to a re-training algorithm configured to re-train the machine learning model in real-time to generate a re-trained machine learning model – providing data is mere data gathering which is insignificant, extra-solution activity. See MPEP 2106.05(g).
The additional elements do not integrate the abstract idea into a practical application.
Step 2B: Does the claim recite additional elements that amount to significantly more than the judicial exception?
The additional elements of:
A method for re-training a machine learning model in real-time – machine learning models recited at a high level are construed as generic models used to implement the abstract idea. See MPEP 2106.05(f)(1).
generating an embedding of an item of data associated with a user of a software application – generating an embedding is merely preparing data so that it can be input to a computer. As such, it is insignificant, extra-solution activity. See MPEP 2106.05(g). Extracting data is well-understood, routine, and conventional. See MPEP 2106.05(d)(II)(v).
storing the embedding of the item of data on a data store prior to a trigger event comprising a user action with respect to the item of data – storing data is mere data gathering which is insignificant, extra-solution activity. See MPEP 2106.05(g). Storing data is well understood, routine and conventional. See MPEP 2106.05(d)(II)(iv).
obtaining data indicative of the trigger event during an online session for the user of the software application – obtaining data is mere data gathering which is insignificant, extra-solution activity. See MPEP 2106.05(g). Retrieving data is well-understood, routine and conventional. See MPEP 2106.05(d)(II)(iv).
retrieving the embedding of the item of data from the data store in response to obtaining the data indicative of the trigger event during the online session – retrieving data is mere data gathering which is insignificant, extra-solution activity. See MPEP 2106.05(g). Retrieving data is well-understood, routine and conventional. See MPEP 2106.05(d)(II)(iv).
providing the updated training data to a re-training algorithm configured to re-train the machine learning model in real-time to generate a re-trained machine learning model – providing data is mere data gathering which is insignificant, extra-solution activity. See MPEP 2106.05(g). Transmitting data is well-understood, routine, and conventional. See MPEP 2106.05(d)(II)(i).
The additional elements do not amount to significantly more. Therefore, the claim is not patent eligible.
Independent claims 9 and 19 recite the same relevant limitations and are similarly rejected. Independent claim 9 recites the additional limitations of “A system for re-training a machine learning model in real-time, the system comprising: a memory including computer executable instructions; and a processor configured to execute the computer executable instructions” – computer components recited at a high level are construed as generic components used to implement the abstract idea. See MPEP 2106.05(f)(2). Claim 19 recites the additional elements of “A non-transitory computer readable medium comprising instructions to be executed in a computer system” – computer components recited at a high level are construed as generic components used to implement the abstract idea. See MPEP 2106.05(f)(2).
The limitations do not integrate the abstract idea into a practical application, nor do they amount to significantly more. Therefore, the independent claims are not patent eligible.
A similar analysis applies to the dependent claims.
Claims 2 and 10 recite the additional elements of “generate the embedding during the online session for the user of the software application” – preparing data to be input to a computer is insignificant, extra-solution activity. See MPEP 2106.05(g). Extracting data is well-understood, routine and conventional. See MPEP 2106.05(d)(II)(v).
Claims 3, 11, and 20 recite the additional elements of “the embedding of the item of data is generated before the online session for the user of the software application” – description of the data merely identifies a technology or field of use. See MPEP 2106.05(h).
Claims 4 and 12 recite the additional elements of “provide the embedding of the item of data as a feature for the re-training algorithm” -inputting data is insignificant, extra-solution activity. Transmitting data is well-understood, routine, and conventional. See MPEP 2106.05(d)(II)(i).
Claim 5 and 13 recite the additional elements of “obtain one or more predictions generated by the re-trained machine learning model, wherein the one or more predictions comprise a predicted label for an unlabeled item of data associated with the user” – obtaining data is insignificant, extra-solution activity. See MPEP 2106.05(g). Transmitting data is well-understood, routine and conventional. See MPEP 21006.05(d)(II)(i).
Claims 6 and 14 recite the additional elements of “provide an embedding generated for the unlabeled item of data before the online session or during the online session as a feature to the re-trained machine learning model” – providing data is insignificant, extra-solution activity. See MPEP 2106.05(g). Extracting data is well-understood, routine and conventional. See MPEP 2106.05(d)(II)(v); and “obtain a predicted label for the unlabeled item of data as an output of the re-trained machine learning model during the online session” – outputting data is insignificant, extra-solution activity. See MPEP 2106.05(g). Transmitting data is well-understood, routine and conventional. See MPEP 2106.05(d)(II)(i).
Claims 7 and 15 recite the additional elements of “update a user interface of the software application during the online session to display the predicted label generated by the re-trained machine learning model” – updating a user interface by displaying data is outputting data, which is insignificant, extra-solution activity. Readjusting is well-understood, routine, and conventional. See MPEP 2106.05(d)(II)(ii).
Claim 8 recites the additional elements of “providing the updated training data to a regression algorithm” – inputting data to an algorithm is insignificant, extra-solution activity. See MPEP 2106.05(g). Transmitting data is well-understood, routine and conventional. See MPEP 2106.05(d)(II)(i). “re-train the machine learning model in real-time to generate the re-trained machine learning model” – re-training a model without a description of the training, or the model is mere instructions to apply data to a black box. See MPEP 2106.05(f)(3).
Claim 16 recites the additional elements of “the re-training algorithm comprises a regression algorithm” – identifying an algorithm without details of the algorithm, parameters, or hyperparameters merely identifies a technology or field of use. See MPEP 2106.05(h).
Claim 17 recites the additional elements of “the regression algorithm comprises a linear regression algorithm or a logistical regression algorithm” – identifying an algorithm without details of the algorithm, parameters, or hyperparameters merely identifies a technology or field of use. See MPEP 2106.05(h).
Claim 18 recites the additional elements of “the re-training algorithm is configured to re-train the machine learning model is less than 200 milliseconds” – applying a constraint or performance measure on an algorithm without a description of how it is accomplished is merely describing a black box with a performance. See MPEP 2106.05(f)(3).
The additional elements do not integrate the abstract idea into a practical application. Nor do they amount to significantly more. Therefore, claims 1-20 are not patent eligible.
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 1-4, 8-12, and 16-20 are rejected under 35 U.S.C. § 103 as being unpatentable over Raj, et al (Edge MLOps: An Automation Framework for AIoT Applications, herein Raj), Xu, et al (DeepType: On-Device Deep Learning for Input Personalization Service with Minimal Privacy Concern, herein Xu), and Wu, et al (DeltaGrad: Rapid retraining of machine learning models, herein Wu).
Regarding claim 1,
Raj teaches A method for re-training a machine learning model in real-time (Raj, Fig. 1, and, page 193, column 1, paragraph2, line 8 “We use the ML pipeline to train and retrain ML models.” And, page 194, column 1, paragraph 5, line 3 “A periodic trigger from the CI-CD process in the Cloud Orchestration layer is implemented to invoke the monitoring feature in the edge devices, to evaluate model drift and to replace or retrain the existing ML model with an alternative [31]. With this approach, the whole process of ML inference at the edge is automated in real-time.”
PNG
media_image1.png
506
798
media_image1.png
Greyscale
In other words, pipeline for training and retraining machine learning models is a method for retraining a machine learning model, and automated in real time is in real-time.) , comprising
generating an embedding of an item of data associated with a user of a software application (Raj, Fig. 1. In other words, data ingestion is generating an embedding of an item of data associated with a user.);
storing the embedding of the item of data on a data store prior to a trigger event comprising a user action with respect to the item of data (Raj, page 193, column 1, paragraph 2, line 3 “Such a service provides computational resources and data storage on-demand to enable training of machine learning models.” In other words, data storage on-demand to enable training is storing the embedding of the item of data.) ;
obtaining data indicative of the trigger event (Raj, Fig. 1, and page 194, column 1, paragraph 6, line 1 “Machine Learning inference and monitoring are automated as part of Continuous Deployment operations. A periodic trigger from the CI-CD process in the Cloud Orchestration layer is implemented to invoke the monitoring feature in the edge devices, to evaluate model drift and to replace or retrain the existing ML model with an alternative [31].” In other words, trigger is obtaining data indicative of a trigger event.)
[during an online session for the user of the software application];
retrieving the embedding of the item of data from the data store in response to obtaining the data indicative of the trigger event (Raj, page 195, column 2, paragraph 4, line 1 “This process enables and maintains continuous integration of sensor to edge by fetching data in real time. Received data is pre-processed, cleaned and formatted for ML inference.” In other words, fetching data is retrieving the embedding of the item of data from the data store.)
[during the online session];
generating updated training data for the machine learning model based, at least in part, on the embedding of the item of data and the user action with respect to the item of data (Raj, Fig. 1, and, page 193, column 1, paragraph 3, line 1 “The ingestion script procures data (based on parameters) and versions the data which will be used for ML model training. As a result of this step, any experiment (i.e. model training or re-training) can be audited and traced back.” In other words, procures data…..for re-training is updated training data.) ; and
providing the updated training data to a [re-training algorithm] configured to re-train the machine learning model in real-time to generate a re-trained machine learning model (Raj, Fig. 1, page 193, column 1, paragraph 4, line 1 “Model Training and retraining: a script that performs all the traditional steps in ML such as data pre-processing, feature engineering, feature scaling before training or retraining any model. Usually, ML models have a set of hyperparameters to tune for fitting the model to the dataset (training set). This step can be done manually, but efficient and automatic solutions such as GridSearch or Random-Search [29], exist.” In other words, data preprocessing…before… retraining any model is providing the updated training data…configured to re-training the machine learning model to generate a re-trained machine learning model.)
Thus far, Raj does not explicitly teach during an online session for the user of the software application and during an online session .
Xu teaches during an online session for the user of the software application and during an online session (Xu, Fig. 3, and, page 197, paragraph 2, subparagraph (3), line 1 “ On-device online training. The deployed personal model will be continuously reinforced during the user input procedure. The training data at this phase is generated and used on-the-fly.”
PNG
media_image2.png
298
918
media_image2.png
Greyscale
In other words, online training is during an online session, and during the user input procedure is during an online session for the user of the software application.)
Both Raj and Xu are directed to training machine learning models, among other things. Raj teaches a method for re-training a machine learning model in real-time, comprising: generating an embedding of an item of data associated with a user of a software application;
storing the embedding of the item of data on a data store prior to a trigger event comprising a user action with respect to the item of data; obtaining data indicative of the trigger event
In view of the teaching of Raj, it would be obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teaching of Xu into Raj. This would result in a method for re-training a machine learning model in real-time, comprising: generating an embedding of an item of data associated with a user of a software application;
storing the embedding of the item of data on a data store prior to a trigger event comprising a user action with respect to the item of data; obtaining data indicative of the trigger event during an online session for the user of the software application; retrieving the embedding of the item of data from the data store in response to obtaining the data indicative of the trigger event during the online session.
One of ordinary skill in the art would be motivated to do this to speed up training and improve user experience. (Xu, page 197:2, paragraph 1, line 1 “An unprecedented large amount of textual content is being generated by the Internet users. Everyday there are 2 million comments posted to Reddit [10], 500 million Tweets [18], 3.5 billion Google search queries [9], at least 100 billion instant messages [3, 7], and 200 billion emails sent and received [1]. A majority of these texts are typed in, and a significant proportion is typed in from mobile devices. No accurate statistics can be found on how much time a mobile device user spends on text input; but no doubt it must be a surprisingly large number. Any instrument that successfully reduces this tedious effort could bring a significantly improved user experience.”)
Thus far, the combination of Raj and Xu does not explicitly teach a re-training algorithm.
Wu teaches a re-training algorithm (Wu, Figure 1, Algorithm 1, and, abstract, line 7 “ To address this problem, we propose the DeltaGrad algorithm for rapid retraining machine learning models based on information cached during the training phase.”
PNG
media_image3.png
364
452
media_image3.png
Greyscale
PNG
media_image4.png
724
476
media_image4.png
Greyscale
In other words, DeltaGrad algorithm for rapid retraining machine learning models is re-training algorithm.)
Both Xu, and the combination of Raj and Xu are directed to retraining machine learning models, among other things. The combination of Raj and Xu teaches a method for re-training a machine learning model in real-time, comprising: generating an embedding of an item of data associated with a user of a software application; storing the embedding of the item of data on a data store prior to a trigger event comprising a user action with respect to the item of data; obtaining data indicative of the trigger event during an online session for the user of the software application; retrieving the embedding of the item of data from the data store in response to obtaining the data indicative of the trigger event during the online session; generating updated training data for the machine learning model based, at least in part, on the embedding of the item of data and the user action with respect to the item of data; and providing the updated training data to
In view of the teaching of The combination of Raj and Xu, it would be obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teaching of Wu into The combination of Raj and Xu. This would result in a method for re-training a machine learning model in real-time, comprising: generating an embedding of an item of data associated with a user of a software application; storing the embedding of the item of data on a data store prior to a trigger event comprising a user action with respect to the item of data; obtaining data indicative of the trigger event during an online session for the user of the software application; retrieving the embedding of the item of data from the data store in response to obtaining the data indicative of the trigger event during the online session; generating updated training data for the machine learning model based, at least in part, on the embedding of the item of data and the user action with respect to the item of data; and providing the updated training data to a re-training algorithm configured to re-train the machine learning model in real-time to generate a re-trained machine learning model.
One of ordinary skill in the art would be motivated to do this because machine learning models change over time and it is expensive to retrain them. (Wu, abstract, line 1 “Machine learning models are not static and may need to be retrained on slightly changed datasets, for instance, with the addition or deletion of a set of datapoints. This has many applications, includ-ing privacy, robustness, bias reduction, and un-certainty quantification. However, it is expensive to retrain models from scratch.”)
Regarding claim 2,
The combination of Raj, Xu, and Wu teaches the method of Claim 1, wherein generating the embedding of the item of data comprises
generating the embedding during the online session for the user of the software application (Xu, Fig. 3, and page 11, paragraph 4, line 3 “In other words, the learned weights of word embeddings and LSTM blocks in the global model are representative enough for input patterns, so that we can focus on the fine-tuning in terms of only the weights of last softmax layer.” In other words, embeddings is embedding, and from Fig. 3, online training is during the online session.)
Regarding claim 3,
The combination of Raj, Xu, and Wu teaches the method of Claim 1, wherein generating the embedding of the item of data comprises
generating embedding before the online session for the user of the software application (Xu, Fig. 3, In other words, offline training is generating embedding before the online session.).
Regarding claim 4,
The combination of Raj, Xu, and Wu teaches the method of Claim 1, wherein providing the updated training data to the re-training algorithm comprises
providing the embedding of the item of data as a feature for the re-training algorithm (Raj, Fig. 1, page 193, column 1, paragraph 4, line 1 “Model Training and retraining: a script that performs all the traditional steps in ML such as data pre-processing, feature engineering, feature scaling before training or retraining any model.” In other words, feature engineering… before training or retraining is providing the embedding as a feature for re-training the algorithm.) .
Regarding claim 8,
The combination of Raj, Xu, and Wu teaches the method of Claim 1, wherein providing the updated training data to a re-training algorithm comprises
providing the updated training data to a regression algorithm configured to re-train the machine learning model in real-time to generate the re-trained machine learning model (Wu, Figure 1, page 3, column 1, paragraph 2, line 1 “We provide empirical results show-ing the speed and accuracy of DeltaGrad, for addition, removal, and continuous updates, on a number of stan-dard datasets.” In other words, continuous updates…of standard datasets is providing the updated training data, and from Figure 1, DeltaGrad is a regression algorithm configured to re-train the machine learning model.) .
Claims 9-12 are system claims corresponding to method claims 1-4, respectively. Otherwise, they are not patentably distinct. The combination of Raj, Xu, and Wu teaches a system (Wu, page 6, column 1, paragraph 1, line 1 “Machine configuration. All experiments are run over a GPU machine with one Intel(R) Core(TM) i9-9920X CPU with 128 GB DRAM and 4 GeForce 2080 Titan RTX GPUs (each GPU has 10 GB DRAM). We implemented DeltaGrad with PyTorch 1.3 and used one GPU for accelerating the tensor computations.” In other words, GPU machine is a system.) Therefore, claims 9-12 are rejected for the same reasons as claims 1-4, respectively.
Claim 19-20 are non-transitory computer readable medium claims corresponding to method claims 1 and 3, respectively. Otherwise, they are not patentably distinct. The combination of Raj, Xu, and Wu teaches a non-transitory computer readable medium (Wu, page 6, column 1, paragraph 1, line 1 “Machine configuration. All experiments are run over a GPU machine with one Intel(R) Core(TM) i9-9920X CPU with 128 GB DRAM and 4 GeForce 2080 Titan RTX GPUs (each GPU has 10 GB DRAM). We implemented DeltaGrad with PyTorch 1.3 and used one GPU for accelerating the tensor computations.” In other words, 128GB of DRAM is a non-transitory, computer readable medium.) Therefore, claims 19-20 are rejected for the same reasons as claim 1 and 3, respectively.
Regarding claim 16,
The combination of Raj, Xu, and Wu teaches the system of Claim 9, wherein
the re-training algorithm comprises a regression algorithm (Wu, Figure 1, and Algorithm 1. In other words, logistic regression model is the re-training algorithm comprises a regression algorithm. Examiner notes that a logistic regression model is inherently trained by a logistic regression training algorithm.)
Regarding claim 17,
The combination of Raj, Xu, and Wu teaches the system of Claim 16, wherein
the regression algorithm comprises a linear regression algorithm or a logistical regression algorithm (Wu, Figure 1. See above mapping. In other words, logistic regression is logistical regression algorithm.).
Regarding claim 18,
The combination of Raj, Xu, and Wu teaches the system of Claim 9, wherein the re-training algorithm is configured to
re-train the machine learning model is less than 200 milliseconds (Wu, Figure 1. In other words, Figure 1 shows the DeltaGrad algorithm has a run time less than 200 milliseconds.).
Claims 5-6, and 13-14 are rejected under 35 U.S.C. § 103 as being unpatentable over Raj, Xu, Wu, and Yang, et al (Effective Multi-Label Active Learning for Text Classification, herein Yang).
Regarding claim 5,
The combination of Raj, Xu, and Wu teaches the method of Claim 1, further comprising: obtaining one or more predictions generated by the re-trained machine learning model, wherein the one or more predictions comprise
Thus far, the combination of Raj, Xu, and Wu does not explicitly teach a predicted label for an unlabeled item of data associated with the user.
Yang teaches a predicted label for an unlabeled item of data associated with the user (Yang, page 918, column 1, paragraph 2, line 4, “To measure the loss reduction, we use Support Vector Machines (SVM) in terms of version space [21] due to the effectiveness of SVM active learning on text classification. In the original work, the loss is modeled for single-label case, and here we extend it to multi-label case. We also propose an effective method to predict labels for multi-label data.” In other words, predict labels for multi-label data is a predicted label for an unlabeled item of data associate with the user.)
Both Yang, and the combination of Raj, Xu, and Wu are directed to training machine learning models, among other things. The combination of Raj, Xu, and Wu teaches the method of claim 1, but does not explicitly teach a predicted label for an unlabeled item of data associated with the user. Yang teaches a predicted label for an unlabeled item of data associated with the user.
In view of the teaching of the combination of Raj, Xu, and Wu, it would be obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teaching of Yang into the combination of Raj, Xu, and Wu. This would result in the method of claim 1, and predicting a label for an unlabeled item of data associated with the user.
One of ordinary skill in the art would be motivated to do this in order to correctly categorize unlabeled data items and save time and money. (Yang, abstract, line 1 “Labeling text data is quite time-consuming but essential for automatic text classification. Especially, manually creating multiple labels for each document may become impractical when a very large amount of data is needed for training multi-label text classifiers. To minimize the human-labeling efforts, we propose a novel multi-label active learning approach which can reduce the required labeled data without sacrificing the classification accuracy.”).
Regarding claim 6,
The combination of Raj, Xu, Wu, and Yang teaches the method of Claim 5, wherein obtaining the one or more predictions generated by the re-trained machine learning model during the online session comprises:
providing an embedding generated for the unlabeled item of data before the online session or during the online session as a feature to the re-trained machine learning model (Xu, Fig. 3, In other words, offline training is generating embedding before the online session.); and
obtaining a predicted label for the unlabeled item of data as an output of the re-trained machine learning model during the online session (Xu, Fig. 3, and page 11, paragraph 4, line 3 “In other words, the learned weights of word embeddings and LSTM blocks in the global model are representative enough for input patterns, so that we can focus on the fine-tuning in terms of only the weights of last softmax layer.” In other words, embeddings is embedding, and from Fig. 3, online training is during the online session.).
Claims 13-14 are system claims corresponding to method claims 5-6, respectively. Otherwise, they are not patentably distinct. Therefore, claims 13-14 are rejected for the same reasons as claims 5-6, respectively.
Claims 7 and 15 are rejected under 35 U.S.C. § 103 as being unpatentable over Raj, Xu, Wu, Yang, and Hazelwood, et al (Applied Machine Learning at Facebook: A Datacenter Infrastructure Perspective, herein Hazelwood.)
Regarding claim 7,
The combination of Raj, Xu, Wu, and Yang teaches the method of Claim 5, further comprising:
Thus far, the combination of Raj, Xu, Wu, and Yang does not explicitly teach updating a user interface of the software application during the online session to display the one or more predictions generated by the re-trained machine learning model.
Hazelwood teaches updating a user interface of the software application during the online session to display the one or more predictions generated by the re-trained machine learning model (Hazelwood, page 622, column 2, paragraph 3, line 10 “ Flow also has tooling for experiment management and a simple user interface which keeps track of all of the artifacts and metrics generated by each workflow execution or experiment. The user interface makes it simple to compare and manage these experiments.” In other words, user interface is user interface and keeps track of all of the artifacts and metrics is display updated data. Examiner notes that one or more predictions is previously mapped to Yang. See mapping of claim 5.)
Both Hazelwood, and the combination of Raj, Xu, Wu, and Yang are directed to training machine learning models, among other things. The combination of Raj, Xu, Wu, and Yang teach the method of claim 5, but does not explicitly teach updating a user interface of the software application during the online session to display the one or more predictions generated by the re-trained machine learning model. Hazelwood teaches updating a user interface of the software application during the online session to display the one or more predictions generated by the re-trained machine learning model.
In view of the teaching of Raj, Xu, Wu, and Yang, it would be obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teaching of Hazelwood into the combination of Raj, Xu, Wu, and Yang. This would result in the method of claim 5, and updating a user interface of the software application during the online session to display the one or more predictions generated by the re-trained machine learning model.
One of ordinary skill in the art would be motivated to do this because machine learning workloads are diverse and complicated and providing a user interface gives greater control and easier access to the user. (Hazelwood, abstract, line 1 “Machine learning sits at the core of many essential products and services at Facebook. This paper describes the hardware and software infrastructure that supports machine learning at global scale. Facebook’s machine learning workloads are extremely diverse: services require many different types of models in practice. This diversity has implications at all layers in the system stack. In addition, a sizable fraction of all data stored at Facebook flows through machine learning pipelines, presenting
significant challenges in delivering data to high-performance distributed training flows”).
Claim 15 is a system claim that corresponds to method claim 7. Otherwise, they are not patentably distinct. Therefore, claim 15 is rejected for the same reasons as claim 7.
The prior art made of record is considered pertinent to applicant’s disclosure:
Chowdury, F., “A new approach to real-time training of dynamic neural networks” discloses a training method in which only the outer layer weights are updated; the hidden layer weights are chosen randomly at the beginning of the process, and left unchanged.
Khurana, R., “Fraud Detection in eCommerce Payment Systems: The Role of Predictive AI in Real-Time Transaction Security and Risk Management” discloses a method that integrates both supervised and unsupervised learning techniques for fraud detection in eCommerce payment systems, with a contributing role of AI in relation to data privacy, improvement of customer authentication, and continuous learning with respect to emerging cyber threats.
Rokni, S. “Autonomous Training of Activity Recognition Algorithms in Mobile Sensors: A Transfer Learning Approach in Context-Invariant Views” discloses an approach for automatic retraining of machine learning algorithms in real-time without the need for any labeled training data.
Tambos, C., US 2016/0026931 A1, “System and Method for Providing a Machine Learning Re-Training Trigger” discloses a system and method that records the important words lists according to a previous naive Bayes classifier for each category. If a new document provides different important words to distinguish the category from other categories, the method would then re-train the system.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to BART RYLANDER whose telephone number is (571)272-8359. The examiner can normally be reached Monday - Thursday 8:00 to 5:30.
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, Miranda Huang can be reached at 571-270-7092. 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.
/Bart I Rylander/Examiner, Art Unit 2124
Prosecution Insights