DETAILED ACTION
1. The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .2. This action is responsive to the following communication: Amendment filed 7/9/25. This action is made final.
3. Claims 1-15 are pending in the case. Claims 1 are independent claims.
Claim Objections
4. Claims 7, 8 and 12 are objected to as being dependent upon a rejected base claim, but would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims.
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 of this title, 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. Claim 1-6, 9-11 and 13-15 are rejected under 35 U.S.C. 103 as being unpatentable over Wang (US 20260087326) and further in view of Ould-Ahmed-Vall (US 20180315157) in further view of BEN YIZHAK (US 20190357297).
Regarding claim 1, Wang discloses a method for generating output feature map data during operation of neural network processing by a processing unit, the processing unit comprising a plurality of computation resources, the method comprising:
obtaining first, real, data to be processed (see paragraph 0068, and FIG. 5B referring to Real Data);
loading the first data into a set of the plurality of computation resources, causing the set of computation resources to generate a computational result, in a first processing cycle of the processing unit (see FIG. 3 and 4 wherein at least some data is generated and sent to a AI/NN/ML model for computation);
detecting a lack of real data for processing in a second processing cycle of the processing unit, which is subsequent to the first processing cycle (Real data, however, is often incomplete, where the real data consists of missing values. While real data often has missing values, existing synthetic data generation methods include solutions to ‘eliminate’ missing data: either by complete-case analysis (i.e., eliminating samples with missing values) or by impute-and-generate method (i.e., imputing missing values and then using the data). ‘Elimination’ as the only approach to deal with missing data fails to leverage the useful information that missing data captures. This is because, in many practical settings, the missing values are not due to some data-independent (e.g., Missing Completely at Random (MCAR)) mechanism. Instead, missing values are often due to underlying data-dependent mechanisms (e.g., Missing at Random (MAR) and Missing Not at Random (MNAR)) that capture complex situational or environmental interactions. Therefore, the complete-case analysis is unfit for all the real-life situations where missing values in the data are due to MAR and MNAR mechanisms, see paragraph 0023, see also FIG. 4A and 4B wherein the synthetic data is after the “real data”);
obtaining second, artificial, data (see at least FIG. 4B, obtaining labeled synthetic data);
loading the second data into an artificially activated set, of the set of computation resources, in the second processing cycle (see at least FIG. 3, wherein synthetic second data is used in a further processing cycle in the AI/NN/ML);
Wang does not disclose inhibiting the second data from affecting the output feature map data; and generating the output feature map data based at least in part on the computational result.
However, Ould-Ahmed-Vall discloses wherein Feature map data output from a neural layer can be based on a set of data output from a set of functions In such embodiment a specific set of low precision instructions are provided that enable automatic adjustment of N-bit neural network data to prevent catastrophic loss of precision. Exemplary transformations or normalizations that may be performed by the normalize and transform logic 2106A-2016B include weight normalization to a range of values or a set of persistent and reversible feature data transformation. In one embodiment, weight normalization can be performed to compress the dynamic range of a set of filter weights to within a predetermined range. Weight data can be normalized, for example, within a range of [−1,1], which can preserve the relative differences between weight values while reducing the overall magnitude of the weight values. In one embodiment, neural network weight or feature map data can be normalized via the mean value of the dataset (paragraph 0222).
The combination of Wang and Ould-Ahmed-Vall would have resulted in the synthetic data learning method to further incorporate Ould-Ahmed-Vall’s teachings of adjusting the datasets to weight their incorporation. One would have been motivated to have combined the teachings because a user in Wang is already involved in using different data sets to incorporate better sets to use for the learning models and determining which of those are active would have made said sets more robust. As such, the combination of teachings would have been obvious to one of ordinary skill in the art as the resulting combination would have been predictable.
Wang does not disclose wherein in the second processing cycle, thereby preventing the artificially activated set of computation resources from being in an idle state in the second processing cycle.
However, BEN YIZHAK discloses wherein As explained herein, each UE 130 comprises a modem 210, a data repository 200 and a processing resource 220. As mentioned, modem 210 is capable of connecting UE 130 to one or more cellular networks 110, at least one of them is a UMTS type cellular network 110. The modem 210 of the UE 130 may be in one of the following non-limiting states: an idle state, one or more intermediate states, and a dedicated channel state. In the idle state the modem 210 is not connected to the cellular network 110. In the dedicated channel state, the modem 210 is connected to the cellular network 110 via a dedicated communication channel. In each of the intermediate states the modem 210 is not connected to the cellular network 110 via the dedicated communication channel and the time-period required to transition from one of the corresponding intermediate states to the dedicated channel state is shorter than the time-period required to transition from the idle state to the dedicated channel state. The processing resource 220 of the UE 130 is configured to maintain a state transitioning regime of the UE 130, wherein the state transitioning regime includes preventing at least one transition of the UE 130 to an idle state for at least one time-period. A further explanation is provided herein, inter alia with reference to FIG. 3 (see paragraph 0060).
The combination of Wang and BEN YIZHAK would have resulted in the synthetic data learning method to further incorporate BEN YIZHAK’s teachings of preventing the resources from being an idle state. One would have been motivated to have combined the teachings because a user in Wang is already involved in using different data sets to incorporate better sets to use for the learning models and determining which of those are active would have made said sets more robust. As such, the combination of teachings would have been obvious to one of ordinary skill in the art as the resulting combination would have been predictable.
Regarding claim 2, Wang discloses wherein the second processing cycle is immediately subsequent to the first processing cycle (see at least FIG. 3 and 4 wherein data is first sent and then synthetic data is sent).
Regarding claim 3, Wang discloses comprising deriving the second data from real data (the data can be real data or synthetic data. In some cases, the real data can include sensitive personal information. The data can be any structured data, such as tabular data, lists, textual data, or temporal data. Tabular data refers to data that is organized in a table with rows and columns. The tabular data can be either numeric data or categorical data. It should be noted that while the data is described as structured data, the data may be any type of data, such as semi-structured data or unstructured data, paragraph 0042).
Regarding claim 4, Wang discloses comprising: obtaining preceding, real, data to be processed before the first data; loading the preceding data into a preceding set, of the set of computation resources, causing the preceding set of computation resources to generate a preceding computational result, in a preceding processing cycle of the processing unit; deriving the second data from the preceding data; and generating the output feature map data based at least in part on the preceding computational result (see FIGS. 3 and 4a/4b wherein the real data is sent first to the synthetic data service before being sent to generate further data. Moreover, the synthetic data is sent back with missing elements to then be sent back to the learning units).
Regarding claim 5, Wang does not disclose comprising maintaining, in the second processing cycle, a deactivated subset of the set of computation resources in an idle state.
However, Ould-Ahmed-Vall discloses wherein Feature map data output from a neural layer can be based on a set of data output from a set of functions In such embodiment a specific set of low precision instructions are provided that enable automatic adjustment of N-bit neural network data to prevent catastrophic loss of precision. Exemplary transformations or normalizations that may be performed by the normalize and transform logic 2106A-2016B include weight normalization to a range of values or a set of persistent and reversible feature data transformation. In one embodiment, weight normalization can be performed to compress the dynamic range of a set of filter weights to within a predetermined range. Weight data can be normalized, for example, within a range of [−1,1], which can preserve the relative differences between weight values while reducing the overall magnitude of the weight values. In one embodiment, neural network weight or feature map data can be normalized via the mean value of the dataset (paragraph 0222).
The combination of Wang and Ould-Ahmed-Vall would have resulted in the synthetic data learning method to further incorporate Ould-Ahmed-Vall’s teachings of adjusting the datasets to weight their incorporation. One would have been motivated to have combined the teachings because a user in Wang is already involved in using different data sets to incorporate better sets to use for the learning models and determining which of those are active would have made said sets more robust. As such, the combination of teachings would have been obvious to one of ordinary skill in the art as the resulting combination would have been predictable.
Regarding claim 6, Wang discloses comprising: obtaining third, artificial, data; loading the third data into a further artificially activated set, of the set of computation resources, in a third processing cycle of the processing unit subsequent to the second processing cycle (see FIGS. 3 and 4a/4b wherein the real data is sent first to the synthetic data service before being sent to generate further data. Moreover, the synthetic data is sent back with missing elements to then be sent back to the learning units).
Wang does not disclose inhibiting the third data from affecting the output feature map data; and maintaining, in the third processing cycle, a further deactivated subset of the set of computation resources in the idle state, wherein the further deactivated subset is larger than the deactivated subset.
Further, Ould-Ahmed-Vall discloses wherein Feature map data output from a neural layer can be based on a set of data output from a set of functions In such embodiment a specific set of low precision instructions are provided that enable automatic adjustment of N-bit neural network data to prevent catastrophic loss of precision. Exemplary transformations or normalizations that may be performed by the normalize and transform logic 2106A-2016B include weight normalization to a range of values or a set of persistent and reversible feature data transformation. In one embodiment, weight normalization can be performed to compress the dynamic range of a set of filter weights to within a predetermined range. Weight data can be normalized, for example, within a range of [−1,1], which can preserve the relative differences between weight values while reducing the overall magnitude of the weight values. In one embodiment, neural network weight or feature map data can be normalized via the mean value of the dataset (paragraph 0222).
Moreover, it would have been obvious in Wang wherein the added data set from FIG. 3 would be larger than the original set as it is adding additional information.
The combination of Wang and Ould-Ahmed-Vall would have resulted in the synthetic data learning method to further incorporate Ould-Ahmed-Vall’s teachings of adjusting the datasets to weight their incorporation. One would have been motivated to have combined the teachings because a user in Wang is already involved in using different data sets to incorporate better sets to use for the learning models and determining which of those are active would have made said sets more robust. As such, the combination of teachings would have been obvious to one of ordinary skill in the art as the resulting combination would have been predictable.
Regarding claim 9, Wang discloses wherein the lack of real data comprises at least one of a lack of real input feature map data and a lack of real weights, and the second, artificial, data comprises artificial input feature map data and artificial weights (Synthetic data is only useful if it is realistic e.g., it mimics the real data and provides similar statistical results. Typically, in industry and in research, the methods of generating synthetic data (or of general data analysis and machine learning) work over the data with no missing values (i.e., complete data). Real data, however, is often incomplete, where the real data consists of missing values. While real data often has missing values, existing synthetic data generation methods include solutions to ‘eliminate’ missing data: either by complete-case analysis (i.e., eliminating samples with missing values) or by impute-and-generate method (i.e., imputing missing values and then using the data). ‘Elimination’ as the only approach to deal with missing data fails to leverage the useful information that missing data captures. This is because, in many practical settings, the missing values are not due to some data-independent (e.g., Missing Completely at Random (MCAR)) mechanism. Instead, missing values are often due to underlying data-dependent mechanisms (e.g., Missing at Random (MAR) and Missing Not at Random (MNAR)) that capture complex situational or environmental interactions. Therefore, the complete-case analysis is unfit for all the real-life situations where missing values in the data are due to MAR and MNAR mechanisms, see paragraph 0023).
Regarding claim 10, Wang does not disclose wherein each computation resource of the plurality of computation resources comprises at least one multiply-accumulate unit, each multiply-accumulate unit configured to multiply a portion of input feature map data by at least one weight.
However, Ould-Ahmed-Vall discloses in one embodiment the operand registers 1414 include three input operands, for example, to support fused multiply-add, multiply-subtract, multiply-accumulate, or related operations. In one embodiment the operand registers 1414 can also store integer values, as in one embodiment the dynamic precision floating point unit supports 32-bit, 16-bit, and 8-bit integer operations. The specific data-type and baseline precision is configurable, in one embodiment, via an input to the control unit 1402 (paragraph 0191).
The combination of Wang and Ould-Ahmed-Vall would have resulted in the synthetic data learning method to further incorporate Ould-Ahmed-Vall’s teachings of adjusting the datasets to weight their incorporation. One would have been motivated to have combined the teachings because a user in Wang is already involved in using different data sets to incorporate better sets to use for the learning models and determining which of those are active would have made said sets more robust. As such, the combination of teachings would have been obvious to one of ordinary skill in the art as the resulting combination would have been predictable.
Regarding claim 11, Wang discloses comprising: causing the artificially activated set to generate an artificial computational result in the second processing cycle (see at least FIG.3 wherein the data is configured/altered to generate artificial data sets).
Wang does not disclose wherein inhibiting the second data from affecting the output feature map data comprises discarding the artificial computational result.
However, Ould-Ahmed-Vall discloses wherein Feature map data output from a neural layer can be based on a set of data output from a set of functions In such embodiment a specific set of low precision instructions are provided that enable automatic adjustment of N-bit neural network data to prevent catastrophic loss of precision. Exemplary transformations or normalizations that may be performed by the normalize and transform logic 2106A-2016B include weight normalization to a range of values or a set of persistent and reversible feature data transformation. In one embodiment, weight normalization can be performed to compress the dynamic range of a set of filter weights to within a predetermined range. Weight data can be normalized, for example, within a range of [−1,1], which can preserve the relative differences between weight values while reducing the overall magnitude of the weight values. In one embodiment, neural network weight or feature map data can be normalized via the mean value of the dataset (paragraph 0222).
The combination of Wang and Ould-Ahmed-Vall would have resulted in the synthetic data learning method to further incorporate Ould-Ahmed-Vall’s teachings of adjusting the datasets to weight their incorporation. One would have been motivated to have combined the teachings because a user in Wang is already involved in using different data sets to incorporate better sets to use for the learning models and determining which of those are active would have made said sets more robust. As such, the combination of teachings would have been obvious to one of ordinary skill in the art as the resulting combination would have been predictable.
Regarding claim 13, Wang discloses comprising: after detecting the lack, lengthening a processing cycle duration of the processor so that a duration of the second processing cycle is greater than a duration of the first processing cycle (as seen FIGS. 3 and 4a/4b it would have been obvious wherein the cycles containing the extra synthetic data would have been longer and more complicated than the raw data).
Regarding claim 14, Wang discloses wherein the processing unit is a neural processing unit (see paragraph 0077, processing systems).
Regarding claim 15, Wang discloses a processing unit configured to perform the method of claim 1 (see paragraph 0077, processing systems).
Response to Amendment
7. Applicant’s arguments with respect to claims X have been considered but are moot in view of the new grounds of rejection.
Conclusion
THIS ACTION IS MADE FINAL. 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 extension fee 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 E CHOI whose telephone number is (571)270-3780. The examiner can normally be reached on M-F: 7-2, 7-10 (PST). If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Michelle Bechtold can be reached on 571-431-0762. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of an application may be obtained from the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see http://pair-direct.uspto.gov. Should you have questions on access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative or access to the automated information system, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/DAVID E CHOI/Primary Examiner, Art Unit 2148
Prosecution Insights