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 § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claim 15 is rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Claim 15 recites “The non-transitory computer readable medium of claim 1”, however claim 1 is a method claim thus this term lacks antecedent basis. The examiner believes that claim 11 was intended to reference claim 15, but was mistakenly written as claim 1.
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, 2, 11, 12 are rejected under 35 U.S.C. 103 as being unpatentable over Wang (20210125070 A1), in view of Dumas (Autoencoder Based Image Compression: Can the learning be quantization independent?) and Daily (US 9020870 B1).
Regarding claim 1, Wang discloses “obtaining a group of descriptors that were outputted by of one or more neural network layers; wherein descriptors of the group of descriptors comprise a first number (N1) of descriptor elements;” (See [0005]; a weight tensor (group of descriptors within a layer) is received from a neural network layer)
“generating a sparse representation of the group of descriptors, wherein the generating comprises:” (See [0005], [0082]; a sparse representation of the group of descriptors is generated through compression)
“repeating, for each training image of multiple training images” (See [0081]; Wang discloses a training process that repeats for each element (training image) of a set)
“outputting, by a neural network, a group of training descriptors related to the training image;” (See [0005]; Wang discloses a neural network outputting a weight tensor (group of training descriptors))
“generating a passive readout unit output, in response to the group of training descriptors;” (See [0005], [0082]; Wang discloses a sparse and lossless output being generated after the weight tensor (group of training descriptors) was outputted. Passive readout is a lossless output)
“adjusting… based on a difference between the group of training descriptors and the decoded output” (See [0107]; Wang discloses measuring distortion by finding the difference between two parameters, such as the difference between an input feature map and an output feature map).
Wang fails to explicitly disclose, “applying a dimension expanding convolution operation on the group of descriptors to provide a group of expanded descriptors; wherein expanded descriptors of the group of expanded descriptors comprises a second number (N2) of expanded descriptor elements, wherein N2 exceeds N1;”
“quantizing the group of expanded descriptors to provide a group of binary descriptors that form a sparse representation of the group of descriptors”
“generating is executed by a readout unit that is trained by a training process”
“decoding the passive readout unit output by a process that reverses the generating of the passive readout unit output, to provide a decoded output;”.
Dumas teaches “applying a dimension expanding convolution operation on the group of descriptors to provide a group of expanded descriptors; wherein expanded descriptors of the group of expanded descriptors comprises a second number (N2) of expanded descriptor elements, wherein N2 exceeds N1;” (See [Page 2-3, Section 3, Paragraph 1]; Dumas discloses applying a dimension expanding convolution operation by using a process that first builds a convolutional autoencoder by using a normal composition of convolutional layers and GDNs, and then reverses each component in the autoencoder by replacing each GDN with an inverse GDN and each convolutional layer with a transpose convolutional layer)
“quantizing the group of expanded descriptors to provide a group of binary descriptors that form a lossless and a sparse representation of the group of descriptors.” (See [Page 4, Section 4, Paragraph 3, 5]; Dumas discloses using a quantization step on each feature map (group of expanded descriptors) of the system and also discloses that the quantization step implements a binarizer to provide a group of binary descriptors)
“decoding the passive readout unit output by a process that reverses the generating of the passive readout unit output, to provide a decoded output;” (See [Page 2-3, Section 3, Paragraph 1]; Dumas discloses decoding the output with a process that reverses the original composition to provide a decoded output).
Wang-Dumas fails to explicitly disclose, “adjusting the passive readout unit…”.
Daily teaches “adjusting the passive readout unit” (See [Page 27, Col 8, Lines 25-27], [Page 28, Column 10, Lines 39-67]; Daily discloses adjusting the readout unit according to the results of a training process).
Therefore, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention having Wang and Dumas before them to modify Wang to --apply a dimension expanding operation on the group of descriptors as well as quantizing the group afterwards. One would be motivated to apply the dimension expanding operation in order to take compressed, low-resolution descriptors and upscale them for better detail, and one would be motivated to quantize the group afterwards to reduce the size of the upscaled group of descriptors for the purpose of higher performance during runtime.
Additionally, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention having Wang, Dumas, and Daily before them to modify Wang to use a readout unit that generates a representation and also to train the readout unit. One would be motivated to do so in order to generate a representation of the group of descriptors for the model and to use the provided training to generate the representation according to the desired specifications.
Additionally, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention having Wang, Dumas, and Daily before them to modify Wang to reverse the passive readout generation to decode the passive readout and then adjusting the output of that reversal based on a difference between the training descriptors group and the output. One would be motivated to do so in order to decode the output to be able to read the results of the passive readout generation and then making adjustments to the output based on the difference between the training descriptors group and the output to correct any missing or flawed parts of the output.
Regarding claim 2, Wang discloses “The applying… comprises independently applying a… process on each descriptor of the group of descriptors” (See [0058]; Wang discloses applying an operation to each pixel (descriptor) in a feature map (group).)
Wang fails to explicitly disclose, “…the dimension expanding convolution operation…”.
Dumas teaches “the dimension expanding convolution operation” (See [Page 2-3, Section 3, Paragraph 1]; Dumas discloses applying a dimension expanding convolution operation).
Therefore, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention having Wang and Dumas before them to modify Wang to use a dimension expanding convolution operation. One would be motivated to do so in order to upscale each descriptor so that each descriptor is in higher detail for better analysis.
Regarding claim 11, this claim is similar in scope to claim 1.
Regarding claim 12, this claim is similar in scope to claim 2.
Claim Rejections - 35 USC § 103
Claims 3, 13 are rejected under 35 U.S.C. 103 as being unpatentable over Wang (20210125070 A1), in view of Dumas (Autoencoder Based Image Compression: Can the learning be quantization independent?) and Daily (US 9020870 B1), and further in view of Zhao (US 20220391676 A1).
Regarding claim 3, Wang-Dumas-Daily fails to explicitly disclose, “quantizing is a top-K quantization”.
Zhao teaches “quantizing is a top-K quantization” (See [0004]; Zhao discloses a neural network that was quantized with top-k quantization).
Therefore, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention having Wang-Dumas-Daily and Zhao before them to modify Wang-Dumas-Daily to use Top-K quantization as one of its preferred quantization techniques. One would be motivated to do so in order to focus on preserving the K most significant descriptors with greater precision while quantizing the remaining descriptors with lower precision for the sake of optimizing performance.
Regarding claim 13, this claim is similar in scope to claim 3.
Claim Rejections - 35 USC § 103
Claims 4, 14 are rejected under 35 U.S.C. 103 as being unpatentable over Wang (20210125070 A1), in view of Dumas (Autoencoder Based Image Compression: Can the learning be quantization independent?) and Daily (US 9020870 B1), and further in view of Kim (Distance-aware Quantization).
Regarding claim 4, Wang-Dumas-Daily fails to explicitly disclose, “quantizing is a argmax quantization”.
Kim teaches “quantizing is a argmax quantization” (See [Page 3, Figure 2]; Kim discloses a quantization technique using a differentiable version of argmax).
Therefore, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention having Wang-Dumas-Daily and Kim before them to modify Wang-Dumas-Daily to use argmax quantization as one of its preferred quantization techniques. One would be motivated to do so in order to quantize the highest probability descriptors that are important to the model according to the argmax function with greater precision while quantizing the remaining descriptors with a lower probability, as this helps with determining which descriptors should receive more compression.
Regarding claim 14, this claim is similar in scope to claim 4.
Claim Rejections - 35 USC § 103
Claim(s) 5, 6, 15, 16 is/are rejected under 35 U.S.C. 103 as being unpatentable over Wang (20210125070 A1), in view of Dumas (Autoencoder Based Image Compression: Can the learning be quantization independent?) and Daily (US 9020870 B1), and further in view of Weisel (US 20220222317 A1).
Regarding claim 5, Wang-Dumas-Daily fails to explicitly disclose, “N2 exceeds N1 by at least a factor of 10”.
Weisel teaches “N2 exceeds N1 by at least a factor of 10” (See [0067]; Weisel discloses that a number of input channels (second group of numbers N2) exceeds a depth of the input data (first group of numbers N1) by at least a factor of 10).
Therefore, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention having Wang-Dumas-Daily and Weisel before them to modify Wang-Dumas-Daily to specify that the second group of numbers N2 would exceed the first group of numbers N1 by a factor of 10. One would be motivated to do so in order to define that N1 needs to be upscaled by at least a factor of 10 to improve the resolution or quality of the input data.
Regarding claim 6, Wang-Dumas-Daily fails to explicitly disclose, “N2 exceeds N1 by at least a factor of 1000”.
Weisel teaches “N2 exceeds N1 by at least a factor of 1000” (See [0067]; Weisel discloses that a number of input channels (second group of numbers N2) exceeds a depth of the input data (first group of numbers N1) by at least a factor of 3 or more, which includes a factor of 1000).
Therefore, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention having Wang-Dumas-Daily and Weisel before them to modify Wang-Dumas-Daily to specify that the second group of numbers N2 would exceed the first group of numbers N1 by a factor of 1000. One would be motivated to do so in order to define that N1 needs to be upscaled by at least a factor of 1000 to improve the resolution or quality of the input data.
Regarding claim 15, this claim is similar in scope to claim 5.
Regarding claim 16, this claim is similar in scope to claim 6.
Claim Rejections - 35 USC § 103
Claims 9, 19 are rejected under 35 U.S.C. 103 as being unpatentable over Wang (US 20210125070 A1), in view of Dumas (Autoencoder Based Image Compression: Can the learning be quantization independent?), and Daily (US 9020870 B1), and further in view of Lipasti (US 20160098629 A1).
Regarding claim 9, Wang-Dumas-Daily fails to explicitly disclose, “training the passive readout circuit by the training circuit”.
Lipasti teaches “training the passive readout circuit by the training circuit” (See [0038]; Lipasti discloses training the readout units).
Therefore, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention having Wang-Dumas-Daily and Lipasti before them to modify Wang-Dumas-Daily to train the passive readout circuit using the training circuit. One would be motivated to do so in order to properly train the passive readout according to the intended training parameters.
Regarding claim 19, this claim is similar in scope to claim 9.
Response to Arguments
The 35 USC 112(a) rejections for claims 1-20 have been withdrawn in view of applicant’s amendments.
The 35 USC 112(b) rejection for claim 15 has not been withdrawn as the applicant has not made amendments regarding this rejection, and the applicant has not argued if it was intended that claim 15 depends on claim 1.
The 35 USC 101 rejections for claims 1-20 have been withdrawn in view of applicant’s amendments.
Applicant’s arguments regarding the 35 USC 103 rejection are moot in view of the new grounds of rejection necessitated by applicant’s amendments.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to DAVID KIM whose telephone number is (571)272-4331. The examiner can normally be reached 7:30 AM - 4:30 PM.
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, Matthew Ell can be reached at (571) 270-3264. 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.
/D.K./ Examiner, Art Unit 2141
/TAN H TRAN/ Primary Examiner, Art Unit 2141
Prosecution Insights