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 .
DETAILED ACTION
This action is responsive to the Application filed on 07/18/2023
Claims 1-20 are pending in the case. Claims 1 and 19-20 are independent claims
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 1-3, 6-7, 10 and 19-20 are rejected under 35 U.S.C. 103 as being unpatentable over "Image compression with a hierarchical neural network", https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=481272, ARAN et al, 01/31/1996, hereinafter referred to as ARAN, in view of BACKHUS et al. (WO 2019177731 A1), hereinafter referred to as BACKHUS and further in view of Jiang et al. (US Patent No.: 11,948,090 B2), hereinafter referred to as Jiang.
With respect to claim 1, ARAN disclose
A signal processing apparatus comprising: one or more processors; and a memory storing instructions which, when the instructions are executed by the one or more processors, cause the signal processing apparatus to function as (On page 1, Image compression with a hierarchical neural network, ARAN does not explicitly recite a "processor", but implicitly requires one. ARAN describes neural network training, neural network inference (compression and reconstruction), and weight calculations and updates, all of which require computational hardware (A neural network cannot run without a processor.) However, if the applicant would like the processing apparatus, memory, processing unit, etc. To be recited explicitly, they can be found in Examiners second presented art BACCHUS which disclose in Fig. 8 and paragraph [0051], a computing system with a processing unit and memory.)
a transfer unit connected with the processing unit and configured to transfer <data> to be stored in a storage unit (On page 2 and Fig. 1, ARAN disclose the system performs the function of storage-transmission (data is sent from one place to another and data is held ).)
wherein the processing unit further executes, on output data outputted from a convolution operation of a first layer among the predetermined layers, an arithmetic operation of a compression layer that is configured by a neural network and compresses data (On page 3 (first paragraph), ARAN discloses compressing the output layer and then it is stored as the compressed image.)
an arithmetic operation of a restoration layer that is configured by a neural network and restores pre-compression data (On page 5 (second paragraph), ARAN involves reconstruction of the overall network, compressing it for storage or transmission, and subsequently restoring it when desired.)
With respect to claim 1, ARAN do not explicitly disclose:
Outputs the first form data to be transmitted to the storage unit, and executes, on the first form data stored in the storage unit
A processing unit configured to execute a convolution operation of predetermined layers constituting a neural network
Outputs input data to be inputted to a convolution operation of a second layer among the predetermined layers
However, it is known by BACKHUS to disclose:
Outputs the first form data to be transmitted to the storage unit, and executes, on the first form data stored in the storage unit (In Fig. 17 and paragraph [0073], BACCHUS discloses quantizing and compressing F filters of the first layer, storing F quantized and compressed filters of the first layer in the memory of the computing system. )
A processing unit configured to execute a convolution operation of predetermined layers constituting a neural network (In Fig. 8 and paragraph [0051], BACCHUS discloses that processor 806 includes a general processor, a central processing unit (CPU). Processor 806 may perform convolution neural network computations, rectified linear unit (reLU) computations, and max pooling computations.)
ARAN and BACKHUS are analogous pieces of art because both references concern a neural network data compression. Accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to modify ARAN, with receiving a large amount of image or text data, compresses it for storage or transmission, and subsequently restores it when desired as taught by ARAN, with quantizing and compressing the parameters of a neural network as taught by BACKHUS. The motivation for doing so would have been to reduce in training and operation (compression-decompression) time (See (Page 2) of ARAN).
With respect to claim 1, ARAN in view of BACKHUS do not explicitly disclose:
Outputs input data to be inputted to a convolution operation of a second layer among the predetermined layers
However, it is known by Jiang to disclose:
Outputs input data to be inputted to a convolution operation of a second layer among the predetermined layers (In Col. 12, lines 22–45, Jiang discloses input is passed through a neural network, producing a feature map (e.g., intermediate output, corresponding to the first layer (convolution output)). A new layer (e.g., SSU layer) is added after the first layer; this layer processes the feature map.)
ARAN in view of BACKHUS and Jiang are analogous pieces of art because both references concern a neural network data compression. Accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to modify Jiang, with compressing a feature map generated by a DNN as taught by Jiang. The motivation for doing so would have been to reduce the quantity of data accessed from a memory when loading the parameters of a neural network into a compute engine (See [0004] of BACKHUS.)
Regarding claim 2, ARAN in view of BACKHUS and Jiang disclose the elements of claim 1. In addition, BACKHUS disclose:
The signal processing apparatus of claim 1, further comprising: the storage unit connected with the transfer unit and configured to store the first form data outputted according to the arithmetic operation of the compression layer (In paragraph [0073], BACKHUS disclose memory component that receives compressed data (from the system) and stores it.)
Regarding claim 3, ARAN in view of BACKHUS and Jiang disclose the elements of claim 1. In addition, Jiang disclose:
The signal processing apparatus of claim 1, wherein the compression layer associated with the convolution operation of the first layer and a compression layer associated with the convolution operation of the second layer are configured to execute the same arithmetic operations (In Col. 12, lines 22-45 Jiang discloses the compression operation step used after the neural network layer and new layer.)
Regarding claim 6, ARAN in view of BACKHUS and Jiang disclose the elements of claim 1. In addition, Jiang disclose:
The signal processing apparatus of claim 1, wherein a neural network including the predetermined layers and a neural network including the compression layer and the restoration layer are configured as separate neural networks (In Col. 12, lines 22–45, Jiang discloses input is passed through a neural network, producing a feature map (e.g., intermediate output, corresponding to the first layer (convolution output)). A new layer (e.g., SSU layer) is added after the first layer; this layer processes the feature map.)
Regarding claim 7, ARAN in view of BACKHUS and Jiang disclose the elements of claim 6. In addition, Jiang disclose:
The signal processing apparatus of claim 6, wherein the compression layer and the restoration layer are trained such that the input data obtained by inputting the first form data outputted from the compression layer to the restoration layer is closer to being the same as the data inputted to the compression layer (In Col. 12, lines 22–45, Jiang discloses input is passed through a neural network, producing a feature map (e.g., intermediate output, corresponding to the first layer (convolution output)). A new layer (e.g., SSU layer) is added after the first layer; this layer processes the feature map.)
Regarding claim 10, ARAN in view of BACKHUS and Jiang disclose the elements of claim 1. In addition, BACKHUS disclose:
The signal processing apparatus of claim 1, further comprising: a transmission unit configured to transmit the first form data outputted according to the arithmetic operation of the compression layer to an apparatus external to the signal processing apparatus (In Fig. 17 and paragraph [0073], BACKHUS disclose quantizing and compressing F filters of the first layer, storing F quantized and compressed filters of the first layer in a memory of the computing system.)
With respect to claim 19, ARAN disclose
Transferring first form data to be stored in a storage unit (On page 2 and Fig. 1, ARAN disclose the system performs the function of storage-transmission (data is sent from one place to another and data is held ).)
wherein in the executing, an arithmetic operation of a compression layer that is configured by a neural network and compresses data is further executed on output data outputted from a convolution operation of a first layer among the predetermined layers (On page 3 (first paragraph), ARAN discloses compressing the output layer and then it is stored as the compressed image.)
and an arithmetic operation of a restoration layer that is configured by a neural network and restores pre-compression data is executed on the first form data stored in the storage unit (On page 5 (second paragraph), ARAN involves reconstruction of the overall network, compressing it for storage or transmission, and subsequently restoring it when desired.)
With respect to claim 19, ARAN do not explicitly disclose:
The first form data to be transmitted to the storage unit is outputted
A method of controlling a signal processing apparatus, the method comprising: executing a convolution operation of predetermined layers constituting a neural network
Input data to be inputted to a convolution operation of a second layer among the predetermined layers is outputted
However, it is known by BACKHUS to disclose:
The first form data to be transmitted to the storage unit is outputted (In Fig. 17 and paragraph [0073], BACCHUS discloses quantizing and compressing F filters of the first layer, storing F quantized and compressed filters of the first layer in the memory of the computing system. )
A method of controlling a signal processing apparatus, the method comprising: executing a convolution operation of predetermined layers constituting a neural network (In Fig. 8 and paragraph [0051], BACCHUS discloses that processor 806 includes a general processor, a central processing unit (CPU). Processor 806 may perform convolution neural network computations, rectified linear unit (reLU) computations, and max pooling computations.)
ARAN and BACKHUS are analogous pieces of art because both references concern a neural network data compression. Accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to modify ARAN, with receiving a large amount of image or text data, compresses it for storage or transmission, and subsequently restores it when desired as taught by ARAN, with quantizing and compressing the parameters of a neural network as taught by BACKHUS. The motivation for doing so would have been to reduce in training and operation (compression-decompression) time (See (Page 2) of ARAN).
With respect to claim 19, ARAN in view of BACKHUS do not explicitly disclose:
Input data to be inputted to a convolution operation of a second layer among the predetermined layers is outputted
However, it is known by Jiang to disclose:
Input data to be inputted to a convolution operation of a second layer among the predetermined layers is outputted (In Col. 12, lines 22–45, Jiang discloses input is passed through a neural network, producing a feature map (e.g., intermediate output, corresponding to the first layer (convolution output)). A new layer (e.g., SSU layer) is added after the first layer; this layer processes the feature map.)
ARAN in view of BACKHUS and Jiang are analogous pieces of art because both references concern a neural network data compression. Accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to modify Jiang, with compressing a feature map generated by a DNN as taught by Jiang. The motivation for doing so would have been to reduce the quantity of data accessed from a memory when loading the parameters of a neural network into a compute engine (See [0004] of BACKHUS.)
With respect to claim 20, ARAN disclose
Transferring first form data to be stored in a storage unit (On page 2 and Fig. 1, ARAN disclose the system performs the function of storage-transmission (data is sent from one place to another and data is held ).)
wherein in the executing, an arithmetic operation of a compression layer that is configured by a neural network and compresses data is further executed on output data outputted from a convolution operation of a first layer among the predetermined layers (On page 3 (first paragraph), ARAN discloses compressing the output layer and then it is stored as the compressed image.)
and an arithmetic operation of a restoration layer that is configured by a neural network and restores pre-compression data is executed on the first form data stored in the storage unit (On page 5 (second paragraph), ARAN involves reconstruction of the overall network, compressing it for storage or transmission, and subsequently restoring it when desired.)
With respect to claim 20, ARAN do not explicitly disclose:
The first form data to be transmitted to the storage unit is outputted
A non-transitory computer-readable storage medium comprising instructions for performing a method of controlling a signal processing apparatus, the method comprising: executing a convolution operation of predetermined layers constituting a neural network
Input data to be inputted to a convolution operation of a second layer among the predetermined layers is outputted
However, it is known by BACKHUS to disclose:
The first form data to be transmitted to the storage unit is outputted (In Fig. 17 and paragraph [0073], BACCHUS discloses quantizing and compressing F filters of the first layer, storing F quantized and compressed filters of the first layer in the memory of the computing system. )
A non-transitory computer-readable storage medium comprising instructions for performing a method of controlling a signal processing apparatus, the method comprising: executing a convolution operation of predetermined layers constituting a neural network (In Fig. 8 and paragraph [0051], BACCHUS discloses that processor 806 includes a general processor, a central processing unit (CPU). Processor 806 may perform convolution neural network computations, rectified linear unit (reLU) computations, and max pooling computations.)
ARAN and BACKHUS are analogous pieces of art because both references concern a neural network data compression. Accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to modify ARAN, with receiving a large amount of image or text data, compresses it for storage or transmission, and subsequently restores it when desired as taught by ARAN, with quantizing and compressing the parameters of a neural network as taught by BACKHUS. The motivation for doing so would have been to reduce in training and operation (compression-decompression) time (See (Page 2) of ARAN).
With respect to claim 20, ARAN in view of BACKHUS do not explicitly disclose:
Input data to be inputted to a convolution operation of a second layer among the predetermined layers is outputted
However, it is known by Jiang to disclose:
Input data to be inputted to a convolution operation of a second layer among the predetermined layers is outputted (In Col. 12, lines 22–45, Jiang discloses input is passed through a neural network, producing a feature map (e.g., intermediate output, corresponding to the first layer (convolution output)). A new layer (e.g., SSU layer) is added after the first layer; this layer processes the feature map.)
ARAN in view of BACKHUS and Jiang are analogous pieces of art because both references concern a neural network data compression. Accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to modify Jiang, with compressing a feature map generated by a DNN as taught by Jiang. The motivation for doing so would have been to reduce the quantity of data accessed from a memory when loading the parameters of a neural network into a compute engine (See [0004] of BACKHUS.)
Claim 4 is rejected under 35 U.S.C. 103 as being unpatentable over ARAN, in view of BACKHUS, Jiang and further in view of Sakaguchi et al. (US Patent No.: 10,817,773 B2), hereinafter referred to as Sakaguchi.
Regarding claim 4, ARAN in view of BACKHUS and Jiang disclose elements of claim 1. ARAN in view of BACKHUS and Jiang do not explicitly disclose:
The signal processing apparatus of claim 1, wherein the compression layer associated with the convolution operation of the first layer and a compression layer associated with the convolution operation of the second layer are configured to execute different arithmetic operations
However, Sakaguchi disclose the limitation (In Col. 16, lines 29–39, Sakaguchi discloses different arithmetic units on the pieces of data (the input data and the weighting coefficient).)
Accordingly, it would have been obvious to a person having ordinary skills in the art before the effective filling date of the claimed invention, having the teaching of ARAN in view of BACKHUS and Jiang before them to include Sakaguchi’s arithmetic processing device to lower the power consumption in performing more reliable arithmetic operations of a neural network (See (Col. 1, lines 41-43) of Sakaguchi)
Claim 5 is rejected under 35 U.S.C. 103 as being unpatentable over ARAN, in view of BACKHUS, Jiang and further in view of Desoli et al. (US Patent No.: 11,593,609 B2), hereinafter referred to as Desoli.
Regarding claim 5, ARAN in view of BACKHUS and Jiang disclose elements of claim 1. ARAN in view of BACKHUS and Jiang do not explicitly disclose:
The signal processing apparatus of claim 1, wherein the processing unit is configured by a plurality of processing units, a first processing unit among the plurality of processing units executes the arithmetic operation of the compression layer and the restoration layer, and a second processing unit among the plurality of processing units executes the convolution operation of the predetermined layers
However, Desoli disclose the limitation (In Col.5-6, lines 66–2, Desoli discloses multiple processing units (first and second convolution accelerators). The first unit performs decompression, decodes compressed data and the convolution accelerator (another hardware block) receives decompressed data.)
Accordingly, it would have been obvious to a person having ordinary skills in the art before the effective filling date of the claimed invention, having the teaching of ARAN in view of BACKHUS and Jiang before them to include Desoli’s , with decompression unit decompresses encoded kernel data in real time during operation of convolutional neural network. The motivation for doing so would have been to improve the efficiency of convolution layers (See (In Col. 4, lines 16-19) of Desoli.)
Claims 8-9 are rejected under 35 U.S.C. 103 as being unpatentable over ARAN, in view of BACKHUS, Jiang and further in view of SHESHKUS et al. (Pub No.: 20220122267 A1), hereinafter referred to as SHESHKUS.
Regarding claim 8, ARAN in view of BACKHUS and Jiang disclose elements of claim 1. ARAN in view of BACKHUS and Jiang do not explicitly disclose:
The signal processing apparatus of claim 1, wherein the compression layer, the restoration layer, and the predetermined layers are included in a single neural network, and the first layer, the compression layer, the restoration layer, and the second layer are configured to be arranged in that order
However, SHESHKUS disclose the limitation (In Fig. 2 and paragraph [0048], SHESHKUS discloses layers arranged so that a first subset precedes a transform layer, and another subset follows an inverse transform layer.)
Accordingly, it would have been obvious to a person having ordinary skills in the art before the effective filling date of the claimed invention, having the teaching of ARAN in view of BACKHUS and Jiang before them to include SHESHKUS’s, with input image to perform semantic image segmentation, so as to produce an output image representing a result of the semantic image segmentation as taught by SHESHKUS. The motivation for doing so would have been to improve the neural network's ability to solve the semantic image segmentation task (See [0081] of SHESHKUS.)
Regarding claim 9, ARAN in view of BACKHUS and Jiang disclose elements of claim 8. ARAN in view of BACKHUS and Jiang do not explicitly disclose:
The signal processing apparatus of claim 8, wherein the compression layer and the restoration layer are trained through training of the single neural network in which the first layer, the compression layer, the restoration layer, and the second layer are configured to be arranged in that order
However, SHESHKUS disclose the limitation (In Fig. 2 and paragraph [0048], SHESHKUS discloses layers arranged so that a first subset precedes a transform layer, and another subset follows an inverse transform layer.)
Accordingly, it would have been obvious to a person having ordinary skills in the art before the effective filling date of the claimed invention, having the teaching of ARAN in view of BACKHUS and Jiang before them to include SHESHKUS’s, with input image to perform semantic image segmentation, so as to produce an output image representing a result of the semantic image segmentation as taught by SHESHKUS. The motivation for doing so would have been to improve the neural network's ability to solve the semantic image segmentation task (See [0081] of SHESHKUS.)
Claims 11-13 and 15-16 are rejected under 35 U.S.C. 103 as being unpatentable over ARAN, in view of BACKHUS, Jiang and further in view of Meng et al. (US Patent No.: 11,948,069 B2), hereinafter referred to as Meng.
Regarding claim 11, ARAN in view of BACKHUS and Jiang disclose elements of claim 1. ARAN in view of BACKHUS and Jiang do not explicitly disclose:
The signal processing apparatus of claim 1, further comprising: a compression/decompression unit configured to execute an arithmetic operation of lossless compression on the output data and an arithmetic operation of decompression on the first form data
a selection unit configured to select execution of either the arithmetic operation according to the compression layer and the restoration layer or the arithmetic operation of the lossless compression and the decompression by the compression/decompression unit
wherein the processing unit performs an arithmetic operation on the output data and an arithmetic operation on the first form data according to the selection by the selection unit
However, Meng disclose the limitations:
The signal processing apparatus of claim 1, further comprising: a compression/decompression unit configured to execute an arithmetic operation of lossless compression on the output data and an arithmetic operation of decompression on the first form data (In Col. 4, lines 28-47, Meng discloses a compression module, a lossless compression scheme which provide increased accuracy over lossy compression schemes.)
a selection unit configured to select execution of either the arithmetic operation according to the compression layer and the restoration layer or the arithmetic operation of the lossless compression and the decompression by the compression/decompression unit (In Col. 4, lines 28-47, Meng discloses a selection module arranged to select a compression scheme which will result in the most efficient, the best compression, and/or minimize the size of the metadata values for each block of the input data.)
wherein the processing unit performs an arithmetic operation on the output data and an arithmetic operation on the first form data according to the selection by the selection unit (In Col. 5, lines 43–49, Meng discloses selection modules 150 and compression modules 160 arranged to compress multiple blocks of the input data 110 substantially in parallel. The output module 170 may then be arranged to combine the outputs of the multiple compression modules 160 into a single output for further processing.)
Accordingly, it would have been obvious to a person having ordinary skills in the art before the effective filling date of the claimed invention, having the teaching of ARAN in view of BACKHUS and Jiang before them to include Meng’s, with compression module for applying the selected compression scheme to the corresponding block to produce compressed neural network activation data. The motivation for doing so would have been to increase accuracy over lossy compression schemes (See(Col.4, lines 30-34) of Meng.)
Regarding claim 12, ARAN in view of BACKHUS, Jiang and Meng disclose the elements of claim 11. In addition, Meng disclose:
The signal processing apparatus of claim 11, wherein in a case where a compression ratio by the compression/decompression unit and an amount of data of the output data satisfy a predetermined condition, the selection unit selects the arithmetic operation of the lossless compression and the decompression by the compression/decompression unit (In Col. 4, lines 28–33, Meng discloses that the processor 100 comprises a compression module 160 arranged to apply the selected compression scheme to the block of input data. The compression schemes used by the compression module 160 are lossless compression schemes which provide increased accuracy over lossy compression schemes.)
Regarding claim 13, ARAN in view of BACKHUS, Jiang and Meng disclose the elements of claim 12. In addition, ARAN disclose:
The signal processing apparatus of claim 12, wherein a compression ratio of compression on the output data by the compression layer is higher than a compression ratio of compression on the output data by lossless compression (On page 7 (paragraph 6), ARAN disclose compression ratio increased through a decrease in the number of hidden nodes used in the network to store the compressed data)
Regarding claim 15, ARAN in view of BACKHUS, Jiang and Meng disclose the elements of claim 11. In addition, Meng disclose:
The signal processing apparatus of claim 11, further comprising: a compression ratio calculation unit configured to calculate a compression ratio of the output data from the available memory bandwidth in the storage unit and an amount of output data, wherein the processing unit performs the arithmetic operation of the lossless compression and the decompression by the compression/decompression unit based on the calculated compression ratio (In Col.6, lines 29–43, Meng discloses uncompressed input data 110 to the selection module 150. The selection module uses the metadata to determine the best/most efficient compression scheme to apply to the particular block of input data.)
Regarding claim 16, ARAN in view of BACKHUS, Jiang and Meng disclose the elements of claim 15. In addition, Meng disclose:
The signal processing apparatus of claim 15, further comprising: wherein the compression/decompression unit includes a plurality of compression/decompression units that perform an arithmetic operation with lossless compression of different compression ratios, and wherein the selection unit selects which compression/decompression unit to use based on the calculated compression ratio (In Col.6, lines 29–43, Meng discloses uncompressed input data 110 to the selection module 150. The selection module uses the metadata to determine the best/most efficient compression scheme to apply to the particular block of input data.)
Claim 14 is rejected under 35 U.S.C. 103 as being unpatentable over ARAN, in view of BACKHUS, Jiang, Meng and further in view of Chong et al. (US Patent No.: 10,855,986 B2), hereinafter referred to as Chong.
Regarding claim 14, ARAN in view of BACKHUS, Jiang and Meng disclose elements of claim 1. ARAN in view of BACKHUS, Jiang and Meng do not explicitly disclose:
The signal processing apparatus of claim 11, further comprising: a measuring unit configured to measure an available memory bandwidth in the storage unit, wherein the compression/decompression unit includes a plurality of compression/decompression units that perform an arithmetic operation with lossless compression of different compression ratios, and the selection unit selects which compression/decompression unit to use based on the measured memory bandwidth
However, Chong disclose the limitation (In Col. 16, lines 9–29, Chong discloses that the compressed data can be stored in a storage device or memory. The storage device or memory can be internal to the neural network device or neural network hardware component, or can be external to the device or hardware component.)
Accordingly, it would have been obvious to a person having ordinary skills in the art before the effective filling date of the claimed invention, having the teaching of ARAN in view of BACKHUS, Jiang and Meng before them to include Chong’s with, bandwidth compression for NN systems. The motivation for doing so would have been to perform bandwidth compression for data of neural network systems (See (Col.1, lines 1-4) of Chong)
Claim 17-18 is rejected under 35 U.S.C. 103 as being unpatentable over ARAN, in view of BACKHUS, Jiang and further in view of Konishi et al. (Pub No.: 20230086727 A1), hereinafter referred to as Konishi.
Regarding claim 17, ARAN in view of BACKHUS and Jiang disclose elements of claim 1. ARAN in view of BACKHUS and Jiang do not explicitly disclose:
The signal processing apparatus of claim 1, further comprising: a determination unit configured to determine, for image data inputted to the processing unit, a degree of importance for each feature based on output data obtained by executing a convolution operation for extracting features related to predetermined characteristic components, wherein the processing unit does not output, as the first form data, data related to the feature depending on the determined degree of importance
However, Konishi disclose the limitation (In paragraph [0034], Konishi discloses determining the degree of importance of the unit in the T.sup.th task; for each of a plurality of layers, determining dissimilar tasks from between a first task to a (T−1).sup.th task, the dissimilar tasks are not similar to the T.sup.th task in terms of behaviors in the layer)
Accordingly, it would have been obvious to a person having ordinary skills in the art before the effective filling date of the claimed invention, having the teaching of ARAN in view of BACKHUS, Jiang and Meng before them to include Konishi’s, with suppressing updating of weight parameters of the plurality of units included in the plurality of layers in accordance with importance degrees in the dissimilar tasks determined for each of the plurality of layers. The motivation for doing so would have been to optimize the parameters of each unit for a task using a set of learning data pieces (See [0003] of Konishi.)
Regarding claim 18, ARAN in view of BACKHUS, Jiang and Meng disclose elements of claim 17. In addition, ARAN disclose :
The signal processing apparatus of claim 17, wherein the processing unit changes whether the first form data stored in the storage unit is used depending on the determined degree of important (On page 2 (first paragraph), ARAN disclose the performance of a vector quantizer compression system is highly dependent on the quality of the code book, and operational times can be long if the code book is large.)
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to EVEL HONORE whose telephone number is (703)756-1179. The examiner can normally be reached Monday-Friday 8 a.m. -5:30 p.m.
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, Mariela D Reyes can be reached at (571) 270-1006. 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.
EVEL HONORE
Examiner
Art Unit 2142
/Mariela Reyes/Supervisory Patent Examiner, Art Unit 2142
Prosecution Insights