DETAILED ACTION
This action is made FINAL in response to the amendments filed on 5/13/2025.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1, 2, 5, 6, 11, 12, 13, 16, 17, 20, 21, 26, 28 – 32, and 34 - 37 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
As to claims 1, 2, 16, 17, 31, and 32
Step 2A, Prong One
The claim recites in part:
reshape the first tensor into at least one second tensor
Under the broadest reasonable interpretation, these limitations are process steps that cover mental processes including an observation, evaluation, judgment or opinion that could be performed in the human mind or with the aid of pencil and paper. If a claim, under its broadest reasonable interpretation, covers a mental process but for the recitation of generic computer components, then it falls within the “Mental Process” grouping of abstract ideas. Accordingly, at Step 2A, Prong One, the claim is directed to an abstract idea.
Step 2A, Prong Two
The judicial exception is not integrated into a practical application. In particular, the claim recites the additional elements of:
transmit and/or store the signal
received and/or retrieve a signal
which amounts to extra-solution activity of gathering data for use in the claimed process. As described in MPEP 2106.05(g), limitations that amount to merely adding insignificant extra-solution activity to a judicial exception do not amount to significantly more than the exception itself, and cannot integrate a judicial exception into a practical application.
The claim further recites:
encode said at least one second tensor in a signal using a Low Displacement Rank (LDR) based approximation of said at least one second tensor, said Low Displacement Rank based approximation of said at least one second tensor having a lower dimension than said first tensor
these elements are recited at a high-level of generality and amounts to no more than adding the words “apply it” to the judicial exception. Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea (See MPEP 2106.05(f)). These limitations also amount to extra solution activity because it is a mere nominal or tangential addition to the claim, amounting to mere data output (see MPEP 2106.05(g)).
The claim further recites a device, processor, non-transitory computer readable medium, and non-transitory computer readable medium which are recited at a high-level of generality and amounts to no more than mere instructions to apply the exception using a generic computer component (See MPEP 2106.05(f)).
In addition, the recitation of tensor, deep neural network, Low Displacement Rank (LDR), and dimension amounts to generally linking the use of the judicial exception to a particular environment of field of use (See MPEP 2106.05(h)). As such, the claim does not integrate the judicial exception into a practical application.
Accordingly, at Step 2A, Prong Two, the additional elements individually or in combination do no integrate the judicial exception into a practical application.
Step 2B
\In accordance with Step 2B, the claim does not include additional elements that are sufficient to amount to significantly more that the judicial exception. As discussed above, the additional elements of:
transmit and/or store the signal
received and/or retrieve a signal
are recited at a high level of generality and amounts to extra-solution activity of receiving data i.e. pre-solution activity of gathering data for use in the claimed process. The courts have found limitations directed to obtaining information electronically, recited at a high level of generality, to be well-understood, routine, and conventional (see MPEP 2106.05(d)(II), “receiving or transmitting data over a network”, "electronic record keeping," and "storing and retrieving information in memory").
In accordance with Step 2B, the claim does not include additional elements that are sufficient to amount to significantly more that the judicial exception. The limitations:
encode said at least one second tensor in a signal using a Low Displacement Rank (LDR) based approximation of said at least one second tensor, said Low Displacement Rank based approximation of said at least one second tensor having a lower dimension than said first tensor
are recited at a high-level of generality and amounts to no more than adding the words “apply it” to the judicial exception. These limitations also amount to extra solution activity because it is a mere nominal or tangential addition to the claim, amounting to mere data output (see MPEP 2106.05(g)). The courts have similarly found limitations directed to displaying a result, recited at a high level of generality, to be well-understood, routine, and conventional. See (MPEP 2106.05(d)(II), "presenting offers and gathering statistics.", “determining an estimated outcome and setting a price”).
The a device, processor, non-transitory computer readable medium, and non-transitory computer readable medium which are recited at a high-level of generality and amounts to no more than mere instructions to apply the exception using a generic computer component (See MPEP 2106.05(f)).
The recitation of tensor, deep neural network, Low Displacement Rank (LDR), and dimension amounts to generally linking the use of the judicial exception to a particular environment of field of use (See MPEP 2106.05(h)).
Accordingly, at Step 2B the additional elements individually or in combination do not amount to significantly more than the judicial exception.
As to claim 5, the limitations “said at least one processor being further configured to obtain a plurality of 1-D vectors by vectorizing said first tensor and obtain said at least one second tensor by stacking said vectors as rows or columns of said at least one second tensor” which amounts to extra-solution activity of gathering data for use in the claimed process. As described in MPEP 2106.05(g), limitations that amount to merely adding insignificant extra-solution activity to a judicial exception do not amount to significantly more than the exception itself, and cannot integrate a judicial exception into a practical application. The courts have found limitations directed to obtaining information electronically, recited at a high level of generality, to be well-understood, routine, and conventional (see MPEP 2106.05(d)(II), “receiving or transmitting data over a network”, "electronic record keeping," and "storing and retrieving information in memory").
As to claims 6 and 11, the limitations “said at least one processor further configured to encode in at least one single at least one information representative of a size of said first tensor or said at least one second tensor, a number of input channels of said layer, a number of output channels of said layer, a size of at least one filter of said layer, or a bias vector of said layer” are recited at a high-level of generality and amounts to no more than adding the words “apply it” to the judicial exception. Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea (See MPEP 2106.05(f)). These limitations also amount to extra solution activity because it is a mere nominal or tangential addition to the claim, amounting to mere data output (see MPEP 2106.05(g)). The courts have similarly found limitations directed to displaying a result, recited at a high level of generality, to be well-understood, routine, and conventional. See (MPEP 2106.05(d)(II), "presenting offers and gathering statistics.", “determining an estimated outcome and setting a price”).
As to claims 11, 26, and 36, the limitations
wherein said 1-D vectors have a size f1f2n1, and said at least one second tensor has a size n2 x f1f2n1 where:
n1 is a number of input channels of said layer
n2 is a number of output channels of said layer
f1 x f2 is the size of at least one filter of said layer
are process steps that cover Mathematical Concepts. If a claim, under its broadest reasonable interpretation, covers a mathematical concept, then it falls within the “Mathematical Concepts” grouping of abstract ideas.
As to claims 13, 28, and 34, the limitations “said at least one processor being further configured to encode in at least one signal an information representative of at least one factor rank of said LDR based approximation” are process steps that cover Mathematical Concepts. If a claim, under its broadest reasonable interpretation, covers a mathematical concept, then it falls within the “Mathematical Concepts” grouping of abstract ideas.
As to claims 20 and 34, the limitations “said at least one processor beinf further configured to obtain a plurality of 1-D vectors as rows or columns of said at least one second tensor and obtain said first sensor from said 1-D vectors” are process steps that cover Mathematical Concepts. If a claim, under its broadest reasonable interpretation, covers a mathematical concept, then it falls within the “Mathematical Concepts” grouping of abstract ideas.
As to claim 29, the limitations “wherein at least one of said at least one representative information is decoded at a layer level” amounts to generally linking the use of the judicial exception to a particular environment of field of use (See MPEP 2106.05(h)).
As to claim 30, the limitations “wherein at least one of said at least one representative information is decoded at a DNN level” amounts to generally linking the use of the judicial exception to a particular environment of field of use (See MPEP 2106.05(h)).
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.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claim(s) 1, 2, 5, 6, 11, 12, 13, 16, 17, 20, 21, 26, 28 – 32, and 34 - 37 is/are rejected under 35 U.S.C. 103 as being unpatentable over Morariu et al (US 2020/0175095) in view of Sainath et al US 10,515,307).
As to claim 1, Morariu et al teaches a device for transmitting and/or storing a first tensor of weights of a layer of a deep neural network (paragraph [0040]…a fusion deep learning module 114, which can execute one or more processes that apply a fusion deep learning model, such as a convolutional neural network), the :
reshape the first sensor into at least one second tensor (paragraph [0038]… a four-dimensional tensor A of shape (R, H, W, D) can be reshaped for stacking with a second tensor B (H, W, D)); and
encode said at least one second tensor in a signal using an approximation of said at least one second tensor (paragraph [0078]… The patch size can vary between (1, 1, D) and (H, W, D). Optimizing the size of the patch between (1, 1, D) and(H, W, D) may allow the stochastic gradient descent (SGD) to converge to local minima more quickly than either a pixel (1, 1, D) or channel (H, W, D) level implementation. The SGD is an iterative method for optimizing a differentiable objective function using stochastic approximation of a gradient descent optimization. SCCA 610 loss is applied to the feature maps, such as KS intermediate feature map 506, PD intermediate feature map 512, and ML intermediate feature map 518 before the stacking is performed), said approximation of said at least one second tensor having a lower dimension than said first tensor (paragraph [0037]… the reduction in size of each tensor dimension to include the most significant feature maps can significantly lower the computational costs of performing computations on the tensor); and
transmit and/or store the signal (paragraph [0081]…any suitable computing system or group of computing systems can be used for performing the operations described herein. For example, FIG. 7 depicts an example of the computing system 700. The implementation of computing system 700 could be used for one or more of an electronic document processing application 102, an object recognition engine 112 and a fusion deep learning module 114. In other embodiments, a single computing system 700 having devices similar to those depicted in FIG. 7 (e.g., a processor, a memory, etc.) combines the one or more operations and data stores depicted as separate systems in FIG. 1).
Morariu et al discloses the claimed invention except for the approximation being a Low Displacement Rank based approximation.
It would have been an obvious matter of design choice for the approximation to be a Low Displacement Rank based approximation, since applicant has not disclosed that the approximation being a Low Displacement Rank based approximation solves any stated problems or is for any particular purpose and it appears that the invention would perform equally well any type of approximation known in the art (Saimath et al teaches in column 5, lines 5 – 20… In some embodiments, Low displacement rank corresponds to highly structured matrices such as circulant and Toeplitz matrices and their inverses. High displacement rank matrices can be used to model increasingly unstructured matrices. in some examples, the displacement rank can be used to control the computational complexity, storage requirements, and modeling capacity of for a compression scheme. In some examples, the displacement rank can be tuned based on application requirements).
Claim 2 has similar limitations as claim 1. Therefore, the claim is rejected for the same reasons as above.
As to claim 5, Morariu et al teaches the device wherein said at least one processor being further configured to obtain a plurality of 1-D vectors by vectorizing said first tensor and obtain said at least one second tensor by stacking said vectors as rows or columns of said at least one second tensor (paragraph [0074]…FIG. 6 depicts an example of using spatial canonical correlation analysis on multiple views of feature maps. One or more operations in this example can be used to implement block 208, as described above with respect to FIG. 2. Generally, canonical correlation analysis applies transformations to linear projections of two or more views, in a vector space (e.g., a two-dimensional vectors), that are maximally correlated. An example of canonical correlation analysis between two vector views can be represented by).
As to claim 6, Morariu et al teaches the device wherein said at least one processor further configured to encode in at least one single at least one information representative of
a size of said first tensor or said at least one second tensor (paragraph [0031]… A tensor size of the state feature map can be the number of values contained with the state feature map),
a number of input channels of said layer (paragraph [0060]… the page image decomposition layers 408),
a number of output channels of said layer (paragraph [0060]… an output tensor (i.e., feature map) is determined by the size of the kernel in the convolutional layer and the stride of the convolution),
a size of at least one filter of said layer (paragraph [0060]… A convolutional layer with 200 filters of size (3, 3, 500) acting on a feature map of shape (H, W, 500) produces an output of (H-2, W-2, 200), unless there is padding, in which case the output size might be the same as the input size with 200 as the feature dimensionality: (H, W, 200)), or
a bias vector of said layer (paragraph [0074]…in a vector space (e.g., a two-dimensional vectors), that are maximally correlated).
As to claim 11, Morariu et al in view of Sainath et al discloses the claimed invention except for wherein said 1-D vectors have a size f1f2n1, and said at least one second tensor has a size n2 x f1f2n1 where:
n1 is a number of input channels of said layer
n2 is a number of output channels of said layer
f1 x f2 is the size of at least one filter of said layer
It would have been obvious to one having ordinary skill in the art at the time the invention was made to wherein said 1-D vectors have a size f1f2n1, and said at least one second tensor has a size n2 x f1f2n1, since it has been held that discovering an optimum value of a result effective variable involves only routine skill in the art. In re Boesch, 617 F.2d 272, 205 USPQ 215 (CCPA 1980).
As to claim 13, Saimath et al teaches said at least one processor being further configured to encode in at least one signal an information representative of at least one factor rank of said LDR based approximation (Saimath et al teaches in column 5, lines 5 – 20… In some embodiments, Low displacement rank corresponds to highly structured matrices such as circulant and Toeplitz matrices and their inverses. High displacement rank matrices can be used to model increasingly unstructured matrices. in some examples, the displacement rank can be used to control the computational complexity, storage requirements, and modeling capacity of for a compression scheme. In some examples, the displacement rank can be tuned based on application requirements).
It would have been obvious for said at least one processor being further configured to encode in at least one signal an information representative of at least one factor rank of said LDR based approximation for the same reasons as above.
Claim 16 has similar limitations as claim 1. Therefore, the claim is rejected for the same reasons as above.
Claim 17 has similar limitations as claim 1. Therefore, the claim is rejected for the same reasons as above.
Claim 20 has similar limitations as claim 5. Therefore, the claim is rejected for the same reasons as above.
Claim 21 has similar limitations as claim 6. Therefore, the claim is rejected for the same reasons as above.
Claim 26 has similar limitations as claim 11. Therefore, the claim is rejected for the same reasons as above.
Claim 28 has similar limitations as claim 13. Therefore, the claim is rejected for the same reasons as above.
As to claim 29, Morariu et al teaches the device wherein at least one of said at least one representative information is decoded at layer level (paragraph [0060]…Each of the convolutional blocks 404 and 408 may be a convolutional layer with learnable filters).
As to claim 30, Morariu et al teaches the device wherein at least one of said at least one representative information is decoded at DNN level (paragraph [0062]…the fusion deep learning module 114 processes the stacked feature map 414 (H.sub.ST, W.sub.ST, D′.sub.OR+D.sub.PI) through convolutional blocks 416 to generate pre-aggregation feature map 417 (H′.sub.ST, W′.sub.ST, D′.sub.ST). The pre-aggregation feature map 417 may have the same or different dimensions as stacked feature map 414 depending on the type and number of filters applied by convolutional blocks 416 as described with regard to convolutional blocks 404 and 410 above. In some embodiments, convolutional blocks 416 perform deconvolution operations on the stacked feature map 414).
Claim 31 has similar limitations as claim 1. Therefore, the claim is rejected for the same reasons as above.
Claim 32 has similar limitations as claim 1. Therefore, the claim is rejected for the same reasons as above.
Claim 34 has similar limitations as claim 5. Therefore, the claim is rejected for the same reasons as above.
Claim 35 has similar limitations as claim 6. Therefore, the claim is rejected for the same reasons as above.
Claim 36 has similar limitations as claim 11. Therefore, the claim is rejected for the same reasons as above.
Claim 37 has similar limitations as claim 13. Therefore, the claim is rejected for the same reasons as above.
Response to Arguments
Applicant's arguments filed 5/13/2025 have been fully considered but they are not persuasive.
Claim Rejections - 35 USC § 101
The 101 Rejection still has not been overcome. The claims are abstract and the steps in the claims can be completed with a mental process and/or generic computer components. Additionally, the steps in the claims do not describe an improvement of technology in any way.
The applicant argues:
The Office Action states, essentially, that the claims are directed to non- statutory subject matter because they recite an abstract idea (mental process) without reciting additional elements that integrate the judicial exception into a practical application.
Applicant respectfully submits that independent claim 1 has been amended to recite “...encode said at least one second tensor in a signal using a Low Displacement Rank (LDR) based approximation of said at least one second tensor, said Low Displacement Rank based approximation of said at least one second tensor having a lower dimension than said first tensor; an transmit and/or store the signal.”
Applicant respectfully submits that encoding and transmitting in the manner recited by the amended claim 1 clearly integrates the alleged judicial exception into a practical application of data compression for transmission and/or storage by compressing representation of a weight tensor. See e.g., Application Publication par. [0029]-[0030].
These comments also apply to independent claims 2, 16, and 17, and accordingly to all dependent claims, and Applicant respectfully submits that these claims are thus allowable over the cited references of record for at least the same reasons stated above.
Based on the arguments presented above, withdrawal of the 35 U.S.C. § 101 rejection of claims 1, 2, 5, 6, 11, 12, 13, 16, 17, 20, 21, 26, 28-32, and 34-37 is respectfully requested.
The examiner strongly disagrees as the limitations “transmit and/or store the signal” and “receive and/or retrieve a signal” are just generic functions of generic computer components. Additionally, the limitations “transmit and/or store the signal” or “receive and/or retrieve a signal” which amounts to extra-solution activity of gathering data for use in the claimed process. As described in MPEP 2106.05(g), limitations that amount to merely adding insignificant extra-solution activity to a judicial exception do not amount to significantly more than the exception itself, and cannot integrate a judicial exception into a practical application. The courts have found limitations directed to obtaining information electronically, recited at a high level of generality, to be well-understood, routine, and conventional (see MPEP 2106.05(d)(II), “receiving or transmitting data over a network”, "electronic record keeping," and "storing and retrieving information in memory").
Claim Rejections - 35 USC § 103
The applicant argues:
Further, to the third point, the cited passages of Morariu, at best, generally discuss”...stochastic approximation of a gradient descent optimization...” but do not include any passages that describe “...encod[ing] said at least one second tensor in a signal using ... [an] ... approximation of said at least one second tensor...” General discussion of an approximation of a gradient descent optimization is not sufficient to teach this claim element.
Further, to the fourth point, the Applicant has in fact stated expressly that LDR approximation solves particular stated problems and provides specific advantages. See e.g., par. [(0029]-[0030]. Accordingly, it is inappropriate to ignore, gloss over, or fail to afford patentable weight to the specifically LDR based approximation of claim 1.
Still further, to the fifth point, the cited portions of Saimath do not establish that LDR is a mere design choice, or otherwise support this conclusion. It is clear that the cited passages of Saimath do not discuss arbitrarily choosing one or the other of the described matrices as “...an approximation of said at least one second tensor...” or that the described matrices are generally interchangeable as a design choice. At best, the cited portions of Saimath generally discuss low displacement rank matrices, circulant matrices, Toeplitz matrices, the inverse of these matrices, and high displacement rank matrices in the context of storage requirements, but are otherwise unrelated to the other claim elements and do not teach “...said approximation of said at least one second tensor having a lower dimension than said first tensor...” as in claim 1.
These comments also apply to independent claims 2, 16, and 17, and accordingly to all dependent claims, and Applicant respectfully submits that these claims are thus allowable over the cited references of record for at least the same reasons stated above.
The examiner disagrees. Morariu’s claims are general and broad and the examiner rejects the claims with the broadest reasonable interpretation. The applicant should go into a lot more detail of the approximation limitation. A LDR approximation is simply an optimization algorithm and approximation algorithms are used to find near-optimal solutions to optimization problems, particularly those that are NP-hard. These algorithms do not guarantee the best solution but aim to come as close as possible to the optimal solution within a reasonable amount of time.
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 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 BRANDON S COLE whose telephone number is (571)270-5075. The examiner can normally be reached Mon - Fri 7:30pm - 5pm EST (Alternate Friday's Off).
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, Omar Fernandez can be reached on 571-272-2589. 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.
/BRANDON S COLE/ Primary Examiner, Art Unit 2128