The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Claim Objections
Claim 5 objected to because of the following informalities: "carry out a learning process recited in claim 1" should be "carry out the learning method recited in claim 1". Appropriate correction is required.
Claim 8 objected to because of the following informalities: “carry out a classification process recited in claim 6” should be “carry out the classification method recited in claim 6”. Appropriate correction is required.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claim 3 is rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Claim 3 recites the limitation "a parameter updating process of updating the parameter with use of a gradient of the function" in line 13. There is insufficient antecedent basis for this limitation in the claim: “the function” could refer either to the score function introduced in claim 1, line 3, or to the differentiable function introduced in claim 3, line 12.
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 1-6 and 8 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. The claim(s) recite(s) a mathematical concept.
Regarding claim 1 and analogous claims 4 and 6: The following limitations recite a mathematical concept:
carrying out a learning process of setting a parameter included in a score function for carrying out two-class classification of data
sets the parameter such that, in a square having a horizontal axis representing a false positive rate and a vertical access representing a true positive rate, an area of a region in which the false positive rate is not more than a given threshold is minimized in a region over a receiver operating characteristic, ROC, curve obtained from a training data group.
The following limitations do not integrate into a practical application or recite significantly more:
A learning apparatus, comprising: at least one processor (well-understood, routine, and conventional generic computer, see MPEP 2106.05(f))
Regarding claim 2:
The following limitations recite a mathematical concept:
The learning apparatus according to claim 1
wherein in the learning process, the at least one processor sets the parameter by a hill descending method with use of a differentiable function which approximates the area
Regarding claim 3:
The following claims recite a mathematical concept:
The learning apparatus according to claim 2
(1) a score calculation process of calculating, with use of the score function, (i) a score s+i of each of N+ positive examples x+i included in training data and (ii) a score s-j of each of N- negative examples x-j included in the training data;
(2) a pair creation process of sorting the N+ positive examples x+i in ascending order of scores and the N- negative examples x-j in descending order of scores and then combining p positive examples x+i respectively having bottom p scores with aN- negative examples x-j respectively having top aN- scores to thereby create pxaN- pairs (x+i,x-j) where p is a natural number satisfying s+p < s-1 < s+p.1 and a is the threshold;
(3) a function creation process of creating, with use of the pxaN- pairs (x+i,x- j), the differentiable function which approximates the area; and
(4) a parameter updating process of updating the parameter with use of a gradient of the function.
The following limitations do not integrate into a practical application or recite significantly more:
wherein in the learning process, the at least one processor sets the parameter by repeating the following processes (1) to (4) until a predetermined end condition is satisfied: (insignificant extra-solution activity).
Regarding claim 5 and analogous claim 8:
The following claims recite a mathematical concept:
a learning process recited in claim 1.
a classification process recited in claim 6.
The following limitations do not integrate into a practical application or recite significantly more:
A computer-readable non-transitory storage medium storing therein a learning program for causing a computer to carry out (well-understood, routine, and conventional generic computer, see MPEP 2106.05(f))
Claim Rejections - 35 USC § 102
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 (i.e., changing from AIA to pre-AIA ) 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.
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
Claim(s) 1, 2, 4-6, and 8 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Zhang and Tang (CN113505692A, hereafter referred to as Zhang).
Regarding claim 1 and analogous claims 4 and 6.
Zhang teaches
A learning apparatus, comprising: at least one processor (In paragraph [0104] of Zhang, “This embodiment uses Adam and SGD optimizers from the deep learning framework PyTorch for joint optimization and is trained on two NVIDIA 2080Ti cards [at least one processor]”.)
the at least one processor carrying out a learning process of setting a parameter included in a score function for carrying out two-class classification of data (In paragraph [0051] of Zhang, “The model of this invention no longer needs a softmax output layer for classification, but directly outputs the similarity between two faces [two-class classification of data] for identity determination”. In paragraph [0098] of Zhang, “After the network extracts the image features and embeds them, they are fed into the back end pAUC loss and the entire network is optimized [setting a parameter included in a score function] through backpropagation”.)
wherein in the learning process, the at least one processor sets the parameter such that, in a square having a horizontal axis representing a false positive rate and a vertical representing a true positive rate, an area of a region in which the false positive rate is not more than a given threshold is minimized in a region over a receiver operating characteristic, ROC, curve obtained from a training data group.
In paragraph [0076] of Zhang, “Given a series of θ values, series of {TPR(θ),FPR(θ)} values can be obtained, and the ROC curve can be plotted as shown in figure 1. The grey area in Figure 2 is the defined pAUC, which refers to the area under the ROC curve when the FPR values is in the interval [α,β], where α and β are hyperparameters.” Figure 2 displays the ROC curve, and is shown below.
PNG
media_image1.png
362
512
media_image1.png
Greyscale
In paragraph [0094] of Zhang, “Substituting equation (7) into (5) and transforming the maximization problem into a minimization problem [is minimized over a receiver operating curve], the objective function for pAUC optimization is expressed as equation (8)”
In paragraph [0096] of Zhang, “Use the pAUC optimization function obtained in Step 3 to train the DCNN, and update [the at least one processor sets the parameter] and optimize the entire network through backpropagation”
In paragraph [0064] of Zhang, “Randomly select identities from the training set [obtained from a training data group, and then randomly select two photos from each identity and perform full permutations. This forms a set of containing 2t photos. The 2t photos constitute n sample pairs, including t pairs of positive examples and n pairs of negative examples.”)
Regarding claim 2:
Zhang teaches
The learning apparatus according to claim 1,
wherein in the learning process, the at least one processor sets the parameter by a hill descending method (In paragraph [0104] of Zhang, “This embodiment uses Adam and SGD [a hill descending method] optimizers from the deep learning framework PyTorch for joint optimization and is trained on two NVIDIA 2080Ti cards”.)
with use of a differentiable function which approximates the area. (In paragraph [0076] of Zhang, “Step 3-5: The pUAC optimization objective function [the differentiable function which approximates the area] is then calculated from [creating, with use of the] the set of P and N0.[ the pxαN- pairs (x+I, x-j)].”)
Regarding claim 5 and analogous claim 8:
Zhang teaches:
A computer-readable non-transitory storage medium storing therein a learning program for causing a computer to carry out (In paragraph [0104] of Zhang, “This embodiment uses Adam and SGD optimizers from the deep learning framework PyTorch for joint optimization and is trained on two NVIDIA 2080Ti cards [at least one processor]”.)
a learning process recited in claim 1.
a classification process recited in claim 6.
Claim Rejections - 35 USC § 103
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 (i.e., changing from AIA to pre-AIA ) 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.
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.
Claim 3 is rejected under 35 U.S.C. 103 as being unpatentable over Zhang and further in view of Donmez and Carbonell (Donmez P, Carbonell JG. Active sampling for rank learning via optimizing the area under the ROC curve. InEuropean Conference on Information Retrieval 2009 Apr 6 (pp. 78-89). Berlin, Heidelberg: Springer Berlin Heidelberg, hereafter referred to as Donmez ).
Regarding claim 3:
Zhang teaches
The learning apparatus according to claim 2,
wherein in the learning process, the at least one processor sets the parameter by repeating the following processes (1) to (4) until a predetermined end condition is satisfied: (In paragraph [0104] of Zhang, “The initial learning rate was set to 0.1, then to 0.01 at the 150,000th iteration, then to 0.001 at the 250,000th iteration, and training [repeating the following processes (1) to (4)] was stopped at the 300,000th iteration [until a predetermined end condition is satisfied]. ”
a score calculation policy of calculating, with use of the score function, (i) a score s+i of each N+ positive examples x+I included in training data and (ii) a score s-j of each of N- negative examples x-j included in the training data; (In paragraph [0064] of Zhang, “Randomly select identities from the training set, and then randomly select two photos from each identity and perform full permutations. This forms a set of containing 2t photos. The 2t photos constitute n sample pairs, including t pairs of positive examples [each N+ positive examples x+I included in training data] and n pairs of negative examples [each of N- negative examples x-j included in the training data]”. In paragraph [0067] of Zhang, “Input the sample pairs into the DCNN [the score function], and the DCNN outputs [calculating, with use of the score function] the embedded features [a score s+i and a score s-] of the sample pairs”)
a pair creation process of sorting [..] the N—negative examples x-j in descending order of scores and then combining p positive examples x+I respectively having bottom p scores with αN- negative examples x-j respectively having top αN- scores to thereby create pxαN- pairs (x+I, x-j) where p is a natural number satisfying s+p < s-1 < s+p+1 and α is the threshold;
(In paragraphs [0075] to [0085] of Zhang,
“Where θ [α] is the decision threshold
Given a series of θ values, series of {TPR(θ),FPR(θ)} values can be obtained, and the ROC curve can be plotted as shown in figure 1. The grey area in Figure 2 is the defined pAUC, which refers to the area under the ROC curve when the FPR values is in the interval [α,β], where α and β are hyperparameters.
“To calculate pAUC, construct two sets:
P = {sI 1I =1 | i = 1,2,…,I} (3) [the N- examples x-j]
N = {sj 1j =1 | j = 1,2,…,J} (4) p positive examples x+I
Where I+J=M, sI and sj are the cosine similarities between the same identity and between different identities, respectively;
Steps 3-4: Obtain a new subset N0 from N [the N- examples x-j] by constraining FPR∈[α,β] with [jα/J,jβ/J] [αN- negative examples], where α and β are two hyperparameters. The construction steps are as follows:
Step 3-4-1: Replace [α,β] with [jα/J,jβ/J], where α and β are two integers;
Step 3-4-2: Sort the similarity scores in descending order [sorting… in descending order of scores], where represents all ‘a’s that satisfy condition b, which will be included in the calculation.
Step 3-4-3: Construct a negative example set N0, N0 is composed of the jα-th to jβ-th samples in the set N after similarity scores are sorted in descending order, denoted as N0= {((sk 1k =1 | k = 1,2,…,K}”)
Step 3-5: The pUAC optimization objective function is calculated from the set of P and N0 [combining p positive examples x+I respectively having bottom p scores with αN- negative examples x-j respectively having top αN- scores to thereby create pxαN- pairs (x+I, x-j)]”)
a function creation process of creating, with use of the pxαN- pairs (x+I, x-j), the differentiable function which approximates the area; and (In paragraph [0076] of Zhang, “Step 3-5: The pUAC optimization objective function [the differentiable function which approximates the area] is then calculated from [creating, with use of the] the set of P and N0.[the pxαN- pairs (x+I, x-j)].”)
a parameter updating process of updating the parameter with the use of a gradient of the function. (In paragraph [0096] of Zhang, , “Use the pAUC optimization function [with the use of a gradient of the function] obtained in Step 3 to train the DCNN, and update [updating the parameter] and optimize the entire network through backpropagation”
Zhang fails to teach
sorting the N+ positive examples x+I in ascending order of scores
Donmez teaches
a pair creation process of sorting the N+ positive examples x+I in ascending order of scores (On page 3 of Donmez, “Let the classifier outputs ci[the N+ positive examples] be sorted in ascending order, i.e. the smallest output value is assigned by the lowest rank”.)
Zhang and Donmez are both related to the same field of endeavor (i.e. area under curve (AUC) calculations in machine learning). In view of the teachings of Donmez it would have been obvious for a person of ordinary skill in the art before the effective filing data of the claimed invention to apply the teachings of Donmez to Zhang in order to apply Donmez’s more efficient method of AUC learning in the context of an algorithm using partial AUC (on page 10 of Donmez, “Our method achieves greater learning efficiency with modest computation time in comparison with the other baselines”).
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to ALMA BENNETT THOMPSON whose telephone number is (571)270-1810. The examiner can normally be reached Monday-Thursday 7:30-5:00 EST, alternate Fridays 7:30-4:00.
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, Michael Huntley can be reached at (303) 297-4307. 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.
/ALMA B THOMPSON/Examiner, Art Unit 2129
/MICHAEL J HUNTLEY/Supervisory Patent Examiner, Art Unit 2129
Prosecution Insights