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 Application filed on 01/05/2024. Claims 1-20 are pending in the case.
Information Disclosure Statement
3. As required by MPEP 609 (c), the Applicants’ submission of the Information Disclosure Statement(s) filed on 08/12/2026 is acknowledged by the examiner and the cited references have been considered in the examination of the claims now pending.
Examiner Comments
4. 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
5. 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.
6. Claims 1-20 are rejected under 35 U.S.C. 103 as being unpatentable over GEORGIADIS (Pub. No.: US 20190370667 A1, Pub. Date: 2019-12-05) in view of Singh ( Pub. No.: US 20210357793 A1, Pub. Date 2021-11-18)
Regarding independent Claim 1,
GEORGIADIS teaches the method for generating a sparse representation of a group of neural network features (see GEORGIADIS: Abstract : A system and a method provide lossless compression of an activation map of a neural network.”), the method comprises:
obtaining a group of neural network features (see GEORGIADIS: Fig.2, [0070], “the process starts at 201. At 202, an activation map is received to be encoded. The activation map has been generated at a layer of a neural network is configured to be a tensor of size H×W×C in which H corresponds to the height of the input tensor, W to the width of the input tensor, and C to the number of channels of the input tensor.”); and
generating a lossless and sparse representation of the group of neural network feature (see GEORGIADIS: Fig.1A, [0023], “To facilitate compression, the H×W×C quantized activation map 103 (the group of neural network feature) may be formatted by a formatter 104 into blocks of values, in which each block is referred to herein as “compress units” 105.” [0024], “Each compress unit 105 may be losslessly encoded, or compressed, independently from other compress units by an encoder 106 (sparse representation)- to form a bitstream 107 (generated lossless and sparse representation). Each compress unit 105 may be losslessly encoded, or compressed, using any of a number of compression techniques, referred to herein as “compression modes” or simply “modes.” Example lossless compression modes include, but are not limited to, Exponential-Golomb encoding, Sparse-Exponential-Golomb encoding, Sparse-Exponential-Golomb-RemoveMin encoding,”), wherein the generating comprises:
generating, by the one or more relevant sparse representation generators, one or more relevant sparse outputs (see GEORGIADIS: Fig.1A, [0067], “when a compress unit is compressed, all available compression modes may be run and the compression mode that has generated the shortest bitstream may be selected. The corresponding index for the selected compression mode may be appended as a prefix to the beginning of the bitstream for the particular compress unit and then the resulting bitstream for the compress unit may be added to the bitstream for the entire activation map”)
processing the one or more relevant sparse outputs to provide the lossless and sparse representation of the group of neural network features (see GEORGIADIS: Fig.1A, [0067], “The process may then be repeated for all compress units for the activation map. Each respective compress unit of an activation map may be compressed using a compression mode that is different from the compression mode used for an adjacent, or neighboring, compress unit. In one embodiment, a small number of compression modes, such as two compression modes, may be available to reduce the complexity of compressing the activation maps.”); and
outputting the lossless and sparse representation of the group of neural network features (see GEORGIADIS: Fig.3, [0076], “The activation map 307 is used at layer L of the neural network to compute an output activation map 308. The output activation map 308 is (optionally) quantized at 309 to form a quantized activation map 310. The quantized activation map 310 is formatted at 311 to form compress units 312. The compress units 312 are encoded at 313 to form a bitstream 314, which is stored a memory (not shown) for later use.”)
GEORGIADIS does not teach the system wherein:
determining, by an allocation unit and based on one or more attributes of the group of neural network features, one or more relevant sparse representation generators out of a set of relevant sparse representation generators;
However, Singh teaches the system wherein:
determining (see Singh : Fig. 28, [0187], “As different encode methods have differing efficiency depending on the type of data to be encoded, encode logic can analyze the kernel or feature map data to encode, as shown at 2802. The encode logic can then determine an encode mode based on the data characteristics of the kernel or feature map data, as shown at 2804.”) by an allocation unit (see Singh: Fig.19, [0186], “coder within a DMA controller, such as the encoded 1916 and DMA controller 1906.”) and based on one or more attributes of the group of neural network features (see Singh: Fig.28, [0169], “Metadata for the encoded data indicates the type of encoding format used for the data. In one embodiment, specific encoding formats can be selected for specific types of data, such as kernel data or feature data. In one embodiment, statistical analysis is performed on the data prior to encoding to enable an appropriate encoder to be selected for each block of data.”), one or more relevant sparse representation generators out of a set of relevant sparse representation generators (see Singh: Fig.28, [0170], “For UAV table encoding, a number of unique absolute values for a block of encoded kernel or feature data can be encoded into a header. The specific unique absolute values can then be encoded, followed by an index map that enables each value of the bit stream to be derived from the unique absolute values. In SM encoding mode, only non-zero values in a block are encoded. The number of non-zero values in a sample block is indicated in the header, followed by a significance map indicating a map of the non-zero values within the block.”).
Because both GEORGIADIS and Singh are in the same/similar field of endeavor of training of the neural network, 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 system of GEORGIADIS to include determining, by an allocation unit and based on one or more attributes of the group of neural network features, one or more relevant sparse representation generators out of a set of relevant sparse representation generators as taught by Singh. One would have been motivated to make such a combination to improve computational operation by creating efficient compression and decompression preserves memory bus bandwidth and reduces system memory access power requirements when performing CNN operations. ( see Singh [0039])
Regarding Claim 2,
GEORGIADIS and Singh teach all the limitations of Claim 1. GEORGIADIS and Singh further teaches the system wherein:
the allocation unit is a router.
Regarding Claim 3,
GEORGIADIS and Singh teach all the limitations of Claim 1. GEORGIADIS and Singh further teaches the system wherein:
different relevant sparse representation generators of the set are associated with different situations (see GEORGIADIS: Fig.1A, [0018], “The received tensor may be formatted into smaller blocks that are referred to herein as “compress units.” Compress units may be independently compressed using a variety of different compression modes. The output generated by the encoder is a compressed bitstream. When a compress unit is decompressed, it is reformatted into its original shape as at least part of a tensor of size H×W×C.”)
Regarding Claim 4,
GEORGIADIS and Singh teach all the limitations of Claim 1. GEORGIADIS and Singh further teaches the system wherein:
the processing is concatenating two or more relevant sparse outputs see GEORGIADIS: Fig.1A, [0035], “Concatenate u,f to produce output bitstream”)
Regarding Claim 5,
GEORGIADIS and Singh teach all the limitations of Claim 1. GEORGIADIS and Singh further teaches the system wherein:
the lossless and sparse representation of the group of neural network features consists essentially of one or more outputs of the one or more relevant sparse representation generators (see Singh: Fig.28, [0169], “Metadata for the encoded data indicates the type of encoding format used for the data. In one embodiment, specific encoding formats can be selected for specific types of data, such as kernel data or feature data. In one embodiment, statistical analysis is performed on the data prior to encoding to enable an appropriate encoder to be selected for each block of data.”), and relevant sparse representation generator indicators that identify the one or more relevant sparse representation generators (see Singh: Fig.28, [0170], “For UAV table encoding, a number of unique absolute values for a block of encoded kernel or feature data can be encoded into a header. The specific unique absolute values can then be encoded, followed by an index map that enables each value of the bit stream to be derived from the unique absolute values. In SM encoding mode, only non-zero values in a block are encoded. The number of non-zero values in a sample block is indicated in the header, followed by a significance map indicating a map of the non-zero values within the block.”).
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 system of GEORGIADIS to include the lossless and sparse representation of the group of neural network features consists essentially of one or more outputs of the one or more relevant sparse representation generators as taught by Singh. One would have been motivated to make such a combination to improve computational operation by creating efficient compression and decompression preserves memory bus bandwidth and reduces system memory access power requirements when performing CNN operations. ( see Singh [0039])
Regarding Claim 6,
GEORGIADIS and Singh teach all the limitations of Claim 5. GEORGIADIS and Singh further teaches the system wherein:
the relevant sparse representation generator indicators are routing bits (see GEORGIADIS: Fig.1A, [0022], “the values of the activation map 101 have not been quantized from floating-point numbers to be integers, the non-quantized values of the activation map 101 may be quantized by a quantizer 102 into integer values having any bit width (i.e., 8 bits, 12 bits, 16 bits, etc.) to form a quantized activation map 103. Quantizing by the quantizer 102, if needed, may also be considered to be a way to introduce additional compression, but at the expense of accuracy.”)
Regarding Claim 7,
GEORGIADIS and Singh teach all the limitations of Claim 1. GEORGIADIS and Singh further teaches the system wherein:
the lossless and sparse representation of the group of neural network features comprises one or more outputs of the one or more relevant sparse representation generators and does not include any output of an irrelevant sparse representation generator (see Singh: Fig.25, [0182], “The non-zero values 2236 of the sample are then encoded in order of appearance within the stream. To decode the exemplary bit stream 2410, decoder logic can initialize an output data buffer to zero. As only a small number of non-zero values are found within the bit stream, the decoder logic can reference the coordinates 2234 of the values to determine specifically to place the non-zero values within the decoded stream.”)
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 system of GEORGIADIS to include one or more outputs of the one or more relevant sparse representation generators and does not include any output of an irrelevant sparse representation generator as taught by Singh. One would have been motivated to make such a combination to improve computational operation by creating efficient compression and decompression preserves memory bus bandwidth and reduces system memory access power requirements when performing CNN operations. ( see Singh [0039])
Regarding Claim 8,
GEORGIADIS and Singh teach all the limitations of Claim 1. GEORGIADIS and Singh further teaches the system wherein:
the lossless and sparse representation of the group of neural network features comprises one or more outputs of the one or more relevant sparse representation generators and one or more outputs of one or more irrelevant sparse representation generator (see Singh: Fig.25, [0182], “The non-zero values 2236 of the sample are then encoded in order of appearance within the stream. To decode the exemplary bit stream 2410, decoder logic can initialize an output data buffer to zero. As only a small number of non-zero values are found within the bit stream, the decoder logic can reference the coordinates 2234 of the values to determine specifically to place the non-zero values within the decoded stream.”)
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 system of GEORGIADIS to include one or more outputs of the one or more relevant sparse representation generators and does not include any output of an irrelevant sparse representation generator as taught by Singh. One would have been motivated to make such a combination to improve computational operation by creating efficient compression and decompression preserves memory bus bandwidth and reduces system memory access power requirements when performing CNN operations. ( see Singh [0039])
Regarding independent Claim 9,
Claim 8 is a directed to a method claim and has similar/same claim limitation as Claim 1 and is rejected under the same rationale.
Regarding Claim 10-16,
Claims 10-16 are directed to a non-transitory computer readable medium claim and has similar/same claim limitation as Claim 2-8 respectively and are rejected under the same rationale.
Regarding Claim 17
GEORGIADIS and Singh teach all the limitations of Claim 16. GEORGIADIS and Singh further teaches the system wherein:
a value of an output of an irrelevant sparse representation generator is indicative that the output was generated by an irrelevant sparse representation generator (see Singh: Fig.28, [0170], “UVC encoding mode can be enabled when the number of unique values within a sample block is small and can be stored in a limited number of bits. As an exemplary but non-limiting example, a bit stream sample having only four unique and non-zero values can be encoded using UVC encoding. In ME encoding mode, the mean value for a sample block is encoded, followed by differential for each value from the mean value. ME encoding mode can be enabled when the values to be encoded have a limited dynamic range and are generally clustered around a mean value.”)
Regarding Claim 18
GEORGIADIS and Singh teach all the limitations of Claim 16. GEORGIADIS and Singh further teaches the system wherein:
relevant sparse representation generators of the set of relevant sparse representation generators are arranged in a hierarchical manner (see GEORGIADIS: Fig.1, [0023], “The compress units 105 may include K elements (or values) in a channel-major order in which K>0; a scanline (i.e., each block may be a row of an activation map); or K elements (or values) in a row-major order in which K>0. Other techniques or approaches for forming compress units 105 are also possible. For example, a loading pattern of activation maps for the corresponding neural-network hardware may be used as a basis for a block formatting technique.”)
Regarding Claim 19
GEORGIADIS and Singh teach all the limitations of Claim 16. GEORGIADIS and Singh further teaches the system wherein:
at least one neural network feature is indicative of at least a part of a sensed information unit sensed by a sensor associated with a vehicle (see GEORGIADIS: Fig.1, [0019], “The neural network applications may be used within autonomous vehicles, mobile devices, robots, and/or other low-power devices (such as drones). The techniques disclosed herein reduce memory consumption by a neural network during training and/or as embedded in a dedicated device. The techniques disclosed herein may be implemented on a general-purpose processing device or in a dedicated device.”)
Regarding Claim 20,
GEORGIADIS and Singh teach all the limitations of Claim 16. GEORGIADIS and Singh further teaches the system wherein:
the lossless and sparse representation of the group of neural network features is indicative of at least a part of a sensed information unit sensed by a sensor associated with a vehicle (see GEORGIADIS: Fig.1, [0019], “The techniques disclosed herein may be applied to reduce memory requirements for activation maps of neural networks that are configured to provide applications such as, but not limited to, computer vision (image classification, image segmentation), natural language processing (word-level prediction, speech recognition, and machine translation) and medical imaging. The neural network applications may be used within autonomous vehicles, mobile devices, robots, and/or other low-power devices (such as drones). The techniques disclosed herein reduce memory consumption by a neural network during training and/or as embedded in a dedicated device. The techniques disclosed herein may be implemented on a general-purpose processing device or in a dedicated device.”)
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 20220076095 A1
QIN, Minghai
Title: Multi-level sparse neural networks with dynamic rerouting
Description: Systems and methods for providing a neural network with multiple sparsity levels include sparsifying a matrix associated with the neural network to form a first sparse matrix;
US 20170177812 A1
SJÕLUND; JENS Ola
Title: SYSTEMS AND METHODS FOR OPTIMIZING TREATMENT PLANNING
Description: This disclosure relates generally to radiation therapy or radiotherapy. More specifically, this disclosure relates to systems and methods for developing a statistically optimal radiation therapy treatment plan to be used during radiotherapy.
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
/CESAR B PAULA/Supervisory Patent Examiner, Art Unit 2145
Prosecution Insights