DETAILED ACTION
This Office Action is sent in response to Applicant’s Communication received 6/28/2023 for application number 18/343,353.
Claims 1-22 are pending.
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 § 102
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 the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
Claim(s) 1-2, 4, 7-8, 12, 14-16, 18-22 is/are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Green et al. (US 2022/0391778 A1).
In reference to claim 1, Green discloses a computer-implemented method (para. 0007) for improving accuracy of machine-learning models (para. 0046), the method comprising: obtaining, by one or more computing devices, component data from a donor model, the component data having a first embedding (client devices obtain an embeddings slices from a global model, para. 0052-57, fig. 1); storing the component data in a persistent database, the component data being input data for a plurality of recipient models (global embeddings are stored in repository, para. 0052); receiving, from a first recipient model of the plurality of recipient models, a first request for the first embedding; transmitting, from the persistent database to the first recipient model, the first embedding (a first client device like 102a can request and receive a first embedding slice, para. 0052-57, fig. 1); and processing, using the first recipient model, the first embedding to generate a first recipient output (client updates local model using the received embeddings, para. 0061-64, and then an output, or prediction, can be made using, in part, the received first embedding slice, para. 0094-0101).
In reference to claim 2, Green discloses the computer-implemented method of claim 1, the method further comprising: receiving, from a second recipient model of the plurality of recipient models, a second request for the first embedding (both clients 102a and 102b can request embeddings slice 1, para. 0052-54), wherein the second request is received from the second recipient model at a different time than the first request is received from the first recipient model (Green discloses that a slice can be assigned an identifier, and the embeddings in the slice are dynamically updated as the embeddings are updated over a plurality of iterations, para. 0057, 0114; thus, a second device could request the first slice at a later time / iteration); transmitting, from the persistent database to the second recipient model, the first embedding; and processing, using the second recipient model, the first embedding to generate a second recipient output (client updates local model using the received embeddings, para. 0061-64, and then an output, or prediction, can be made using, in part, the received first embedding slice, para. 0094-0101).
In reference to claim 4, Green discloses the computer-implemented method of claim 2, wherein the second request is received a period of time after the first request is received, and wherein a value associated with the first embedding has changed during the period of time that has elapsed between the first request and the second request (embeddings in the slice are dynamically updated as the embeddings are updated over a plurality of iterations, para. 0057, 0114; thus, a second device could request the first slice at a later time / iteration after the slice has been updated).
In reference to claim 7, Green discloses the computer-implemented method of claim 1, wherein the component data includes embeddings that are not human interpretable, and wherein the first recipient output is a prediction that is human interpretable (global model embeddings would not be human-interpretable, para. 0002-03; the client prediction output is interpretable, like a prediction of a positive entity, para. 0089-91, or items of interest to a user, para. 0097).
In reference to claim 8, Green discloses the computer-implemented method of claim 1, wherein the first embedding is an embedding that is inputted into the first recipient model to generate the first recipient output (client updates local model using the received embeddings as input, para. 0061-64, and then an output, or prediction, can be made using, in part, the received first embedding slice, para. 0094-0101).
In reference to claim 12, Green discloses the computer-implemented method of claim 11, wherein updates to the internal model parameters of the first recipient model are not back propagated to the donor model (internal model embeddings of the first device are not directly used to update the global model: global model can be trained on local training data that has been obscured for privacy, para. 0072-74, and local updates are averaged to update the global model rather than being used directly, para. 0022).
In reference to claim 14, Green discloses the computer-implemented method of claim 1, wherein the component data are embeddings that are outputs of the donor model (global embedding slices come from a global model, para. 0052-57, fig. 1), and wherein the embeddings are continuously transferred to the plurality of recipient models (global model and local models are continuously updated, para. 0071).
In reference to claim 15, Green discloses the computer-implemented method of claim 1, wherein the donor model and the plurality of recipient models are stored as separate model files in different devices (see clients and server in fig. 1).
In reference to claim 16, Green discloses the computer-implemented method of claim 1, wherein the donor model and the plurality of recipient models are run in separate processes (see clients and server in fig. 1) and at different times (global embeddings can be pre-trained, para. 0052, so the global model would run at an earlier time than the client models).
In reference to claim 18, Green discloses the computer-implemented method of claim 1, wherein at least the first recipient model is trained separately from the donor model, and wherein the donor model and the first recipient model are trained in different devices (see clients and server in fig. 1, para. 0052-71).
In reference to claim 19, Green discloses the computer-implemented method of claim 1, further comprising: periodically re-generating, using the donor model, the component data; and replacing the component data in the persistent database with the re-generated component data (global model can periodically or continually updated, para. 0071, and replace embeddings with new embeddings, para. 0057).
In reference to claim 20, this claim is directed to a system associated with the method claimed in claim 1 and is therefore rejected under a similar rationale.
In reference to claim 21, this claim is directed to a system associated with the method claimed in claim 2 and is therefore rejected under a similar rationale.
In reference to claim 22, this claim is directed to a non-transitory computer-readable medium associated with the method claimed in claim 1 and is therefore rejected under a similar rationale.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
Claim(s) 3, 5, 9-11, 13 is/are rejected under 35 U.S.C. 103 as being unpatentable over Green et al. (US 2022/0391778 A1) as applied to claim 1 above, and in further view of Bent, III et al. (US 2020/0402098 A1).
In reference to claim 3, Green does not explicitly teach the computer-implemented method of claim 2, wherein the first recipient model predicts a clickthrough rate of a content item, and wherein the second recipient model predicts a quality value of a landing page associated with the content item (Green teaches models predicting user’s preferences, para. 0096-97, but not these specific predictions).
Bent teaches the computer-implemented method of claim 2, wherein the first recipient model predicts a clickthrough rate of a content item (click-through rate, para. 0062, 70), and wherein the second recipient model predicts a quality value of a landing page associated with the content item (landing page relevance value, para. 0070).
It would have been obvious to one of ordinary skill in art, having the teachings of Green and Bent before the earliest effective filing date, to modify the prediction of Green to include the clickthrough and landing page quality of Bent.
One of ordinary skill in the art would have been motivated to modify the prediction of Green to include the clickthrough and landing page quality of Bent because Green teaches predicting a user’s preferences, para. 0096-97, and Bent teaches user preference measures that are useful for content providers in determining user’s preferences, para. 0002-03.
In reference to claim 5, Green does not explicitly teach the computer-implemented method of claim 1, wherein the first recipient output is a predicted Click-Through Rate (pCTR) that calculates a probability that a content item is clicked when the content item is shown to a user.
Bent teaches the computer-implemented method of claim 1, wherein the first recipient output is a predicted Click-Through Rate (pCTR) that calculates a probability that a content item is clicked when the content item is shown to a user (click-through rate, para. 0062, 70).
It would have been obvious to one of ordinary skill in art, having the teachings of Green and Bent before the earliest effective filing date, to modify the prediction of Green to include the clickthrough rate of Bent.
One of ordinary skill in the art would have been motivated to modify the prediction of Green to include the clickthrough rate of Bent because Green teaches predicting a user’s preferences, para. 0096-97, and Bent teaches user preference measures that are useful for content providers in determining user’s preferences, para. 0002-03.
In reference to claim 9, Green teaches the computer-implemented method of claim 1, wherein the donor model is an online model that learns continuously from … data being fed into the donor model, and wherein the donor model is continuously being trained based on … data (global model is online and continuously updated, para. 0071).
However, Green does not explicitly teach real-time data.
Bent teaches real-time data (para. 0028).
It would have been obvious to one of ordinary skill in art, having the teachings of Green and Bent before the earliest effective filing date, to modify the data of Green to include the real-time data of Bent.
One of ordinary skill in the art would have been motivated to modify the data of Green to include the real-time data of Bent because Green teaches using recent data is desirable, para. 0050, and Bent teaches immediate, real-time data, para. 0027-28.
In reference to claim 10, Green teaches the computer-implemented method of claim 1, wherein the component data is updated based on [recent] data (only data collected within a recent window is used to update model / embeddings, para. 0050), and wherein the first embedding is updated based on a change to a user preference that is captured in the [recent] data (models predicts user’s preferences, para. 0096-97, so if the collected data shows a change in the user preference, it would be reflected in the update to the local model).
However, Green does not explicitly teach real-time data.
Bent teaches real-time data (para. 0028).
It would have been obvious to one of ordinary skill in art, having the teachings of Green and Bent before the earliest effective filing date, to modify the data of Green to include the real-time data of Bent.
One of ordinary skill in the art would have been motivated to modify the data of Green to include the real-time data of Bent because Green teaches using recent data is desirable, para. 0050, and Bent teaches immediate, real-time data, para. 0027-28.
In reference to claim 11, Green teaches the computer-implemented method of claim 1, wherein the plurality of recipient models are online models (online models, para. 0071), and wherein the first recipient model generates the recipient output based on ... data (only data collected within a recent window is used to update model / embeddings, para. 0050; it would be obvious recent data would also be used for predictions, para. 0096-97), and wherein internal model parameters of the first recipient model are updated based on the [recent] data (model updated with recent data, para. 0050).
However, Green does not explicitly teach real-time data.
Bent teaches real-time data (para. 0028).
It would have been obvious to one of ordinary skill in art, having the teachings of Green and Bent before the earliest effective filing date, to modify the data of Green to include the real-time data of Bent.
One of ordinary skill in the art would have been motivated to modify the data of Green to include the real-time data of Bent because Green teaches using recent data is desirable, para. 0050, and Bent teaches immediate, real-time data, para. 0027-28.
In reference to claim 13, Green does not explicitly teach the computer-implemented method of claim 1, wherein the first recipient model serves a prediction based on live internet traffic data.
Bent teaches the computer-implemented method of claim 1, wherein the first recipient model serves a prediction based on live internet traffic data (predictions are for live online content, para. 0027-28).
It would have been obvious to one of ordinary skill in art, having the teachings of Green and Bent before the earliest effective filing date, to modify the predictions of Green to include the live content of Bent.
One of ordinary skill in the art would have been motivated to modify the predictions of Green to include the live content of Bent because Green teaches using recent data is desirable, para. 0050, and Bent teaches making predictions for live content, para. 0027-28.
Claim(s) 6 is/are rejected under 35 U.S.C. 103 as being unpatentable over Green et al. (US 2022/0391778 A1) as applied to claim 1 above, and in further view of Azarbakht et al. (US 2022/0215345 A1).
In reference to claim 6, Green does not explicitly teach the computer-implemented method of claim 1, wherein the donor model is a large model having more than 100 million parameters, and wherein the first recipient model is a small model having less than 100 million parameters.
Azarbakht teaches the computer-implemented method of claim 1, wherein the donor model is a large model having more than 100 million parameters, and wherein the first recipient model is a small model having less than 100 million parameters (knowledge distillation is used to take BERT large with 340 million parameters to BERT-small with 29 million parameters, para. 0099-0104).
It would have been obvious to one of ordinary skill in art, having the teachings of Green and Azarbakht before the earliest effective filing date, to modify the models of Green to include the parameter sizes of Azarbakht.
One of ordinary skill in the art would have been motivated to modify the models of Green to include the parameter sizes of Azarbakht because it would allow for making faster predictions with negligible accuracy loss (Azarbakht, para. 0099-0104).
Claim(s) 17 is/are rejected under 35 U.S.C. 103 as being unpatentable over Green et al. (US 2022/0391778 A1) as applied to claim 1 above, and in further view of Yin et al. (US 2022/0284283 A1).
In reference to claim 17, Green does not explicitly teach the computer-implemented method of claim 1, wherein the receiving, from a first recipient model of the plurality of recipient models, a first request for the first embedding includes: receiving, from the first recipient model, the first request via a remote procedure call (RPC).
Yin teaches the computer-implemented method of claim 1, wherein the receiving, from a first recipient model of the plurality of recipient models, a first request for the first embedding includes: receiving, from the first recipient model, the first request via a remote procedure call (RPC) (para. 0541).
It would have been obvious to one of ordinary skill in art, having the teachings of Green and Yin before the earliest effective filing date, to modify the request of Green to include the RPC of Yin.
One of ordinary skill in the art would have been motivated to modify the request of Green to include the RPC of Yin because Green does not explicitly state how the first client requests embedding slices, and Green provides a standard way of making such a request for data.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. The references cited on the Notice of References Cited and not used above generally teach background information on knowledge distillation and online learning.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Andrew T. Chiusano whose telephone number is (571)272-5231. The examiner can normally be reached M-F, 10am-6pm.
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, Tamara Kyle can be reached at 571-272-4241. 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.
/ANDREW T CHIUSANO/Primary Examiner, Art Unit 2144
Prosecution Insights