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 .
Claims 1-12 have been examined.
Priority
Receipt is acknowledged of certified copies of papers required by 37 CFR 1.55.
Claim Rejections - 35 USC § 102
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.
Claims 1-5 and 7-11 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by “End-to-End Incremental Learning” by Castro et al. (“Castro”).
In regard to claim 1, Castro discloses:
1. A model training method implemented by a processor, comprising: See Castro, Fig. 2, broadly depicting a training method.
obtaining a pre-trained model, See Castro, p. 7, Fig. 2, depicting a training process wherein a first iteration represents an initially trained, i.e. “pre-trained,” model defined in a “representative memory.”
an old dataset, and a new dataset, Castro, p. 7, section 4, “Our training set is composed of samples from the new classes and exemplars from the old classes stored in the representative memory.”
wherein the pre-trained model is a machine-learning model trained by using the old dataset, the old dataset comprises a plurality of old training samples, See Castro, p. 7, Fig. 2, depicting a training process wherein a first iteration represents “pre-training” a model using an initial “old” dataset.
the new dataset comprises a plurality of new training samples, and training of the pre-trained model has not yet used the new dataset; See Castro, p. 7, Fig. 2, depicting a training process wherein a second iteration provides “New samples” which were not present in an initial first iteration.
reducing the old training samples of the old dataset to generate a reduced dataset; and See Castro, p. 7, Fig. 2, e.g. “Representative memory updating.” Also see section 4 at the top of p. 8, e.g. “First, the memory unit removes samples from the stored classes to allocate space for samples from the new classes. Then, the most representative samples from the new classes are selected, and stored in the memory unit according to the selection algorithm.”
using the reduced dataset and the new dataset to tune the pre-trained model. See Castro, p. 7, Fig. 2, e.g. “Balanced fine-tuning.” Also 2nd paragraph of section 4, “Our training set is composed of samples from the new classes and exemplars from the old classes stored in the representative memory.”
In regard to claim 2, Castro also discloses:
2. The model training method as claimed in claim 1, wherein reducing the old training samples in the old dataset comprise: rearranging a sequence of the old training samples; and selecting a portion of the old training samples according to the rearranged sequence of the old training samples. See p. 5, end of section 3.1, e.g. “As the samples are stored in a sorted list, this operation is trivial. The memory unit only needs to remove samples from the end of the sample set of each class.”
In regard to claim 3, Castro also discloses:
3. The model training method as claimed in claim 1, wherein reducing the old training samples of the old dataset comprise: performing clustering on the old training samples to generate at least one group; and selecting a portion from the at least one group. Castro, p. 5, 1st paragraph of section 3.1, “The first setup considers a memory with a limited capacity of K samples. As the capacity of the memory is independent of the number of classes, the more classes stored, the fewer samples retained per class. The number of samples per class, n, is thus given by n = [K/c], where c is the number of classes stored in memory, and K is the memory capacity.” Also see 2nd paragraph of section 3.1, “This is based on herding selection [36], which produces a sorted list of samples of one class based on the distance to the mean sample of that class.”
In regard to claim 4, Castro also discloses:
4. The model training method as claimed in claim 3, wherein the at least one group comprises a first group and a second group, and selecting the portion from the at least one group comprises: selecting old training samples of same quantity from the first group and the second group respectively. Castro, p. 5, 1st paragraph of section 3.1, “The first setup considers a memory with a limited capacity of K samples. As the capacity of the memory is independent of the number of classes, the more classes stored, the fewer samples retained per class. The number of samples per class, n, is thus given by n = [K/c], where c is the number of classes stored in memory, and K is the memory capacity.”
In regard to claim 5, Castro also discloses:
5. The model training method as claimed in claim 1 after tuning the pre-trained model, further comprising: merging the old dataset and the new dataset to generate another old dataset; and associating the another old dataset with the pre-trained model. See Castro, p. 7, Fig. 2, “Representative memory updating.”
In regard to claim 7, Castro discloses:
7. A model training apparatus, comprising: a memory configured to store a program code; and a processor coupled to the memory, executing the program code, and configured to: Castro discloses implementation using “MatConvNet” (see section 5) using the “CIFAR-100” dataset (see section 6), which requires a processor and memory in order to execute the related software.
All further limitations of claim 7 have been addressed in the above rejection of claim 1.
In regard to claims 8-11:
Parent claim 7 is addressed above. All further limitations of claims 8-11 have been addressed in the above rejections of claims 2-5, respectively.
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 6 and 12 are rejected under 35 U.S.C. 103 as being unpatentable over Castro as applied above, and further in view of U.S. Patent Application Publication 20190102700 by Babu et al. ("Babu").
In regard to claim 6, Castro does not expressly disclose:
6. The model training method as claimed in claim 1 after obtaining the pre-trained model, the old dataset, and the new dataset, further comprising: determining whether the old dataset has a label associated with the pre-trained model; and determining the pre-trained model that was trained by using the old dataset according to the label. However, this is taught by Babu. See Babu ¶ 0044, “The different versions of a model might differ in their hyper-parameters or other parameters, and may be trained using different training data. The different versions may be evaluated against various test datasets to identify one or more of them to deploy in a production environment.” It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to use Babu’s version labels in order to identify a particular model for deployment as suggested by Babu.
In regard to claim 12:
Parent claim 7 is addressed above. All further limitations of claim 12 have been addressed in the above rejection of claim 6.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Brownlee, “How to Update Neural Network Models With More Data,” See p. 4 under “Update Model on Old and New Data,” e.g. “A less extreme version would be to use the existing model as a starting point and update it based on the combined dataset.”
U.S. Patent Application Publication 20120284791 by Miller et al. See ¶ 0056, “We have described how to learn the (reference/null) distributions from given training data. In practice, these may change over time. Thus, on-line procedures for estimating these distributions can be applied (i.e., for the models considered here, on-line estimation of mixture models). Some of the data for such adaptation may consist of the samples from the test batch that are not deemed to be anomalous. The sample batch may also change in an on-line fashion, with new samples entering and old samples leaving.”
U.S. Patent Application Publication 20180018590 by Szeto et al. See ¶ 0088, “If there are no discovered or apparent correlations, principle component analysis (PCA) can be leveraged to reduce the dimensionality of the private data 322. Once the dimensionality is reduced, a new trained actual model can be generated and compared to the original trained actual model to ensure that no knowledge has been lost after dimensional reduction and that accuracy in the model is maintained.”
U.S. Patent Application Publication 20200175384 by Zhang et al. See ¶ 0058, “In some embodiments, the system 400 uses parts of the old data to fine tune the existing model 402, along with the new data”
Any inquiry concerning this communication or earlier communications from the examiner should be directed to James D Rutten whose telephone number is (571)272-3703. The examiner can normally be reached M-F 9:00-5:30 ET.
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, Li B Zhen can be reached at (571)272-3768. 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.
/James D. Rutten/Primary Examiner, Art Unit 2121
Prosecution Insights