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 .
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 1-10, 12-18 and 20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Regarding Claim 1,
Claim 1 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 1 is directed to a computer-implemented method of training a neural network, which is directed to a process, one of the statutory categories.
Step 2A Prong One Analysis: The limitations:
“obtaining a probability of a noisy label (P) given x1”
“obtaining a first transition matrix (PTM), wherein the obtaining the PTM comprises including in the PTM a term based on the first output and P”
“combining the PTM with a second transition matrix (NTM) to obtain a third transition matrix (WKM)”
“wherein the updating comprises minimizing a loss function based on the WKM”
As drafted, under their broadest reasonable interpretations, cover mathematical concepts, i.e., mathematical relationships, mathematical formulas or equations, and mathematical calculations. The above limitations in the context of this claim correspond to mathematical relationships and calculations.
Step 2A Prong Two Analysis: The judicial exceptions are not integrated into a practical application. In particular, the claim recites additional elements that are mere instructions to apply (See MPEP 2106.05(f)) and insignificant extra-solution activity (See MPEP 2106.05(g)).
The limitations:
“A computer-implemented method of training a neural network”
“training the neural network using the data set to obtain a first neural network”
“obtaining a first output of the first neural network in response to a first instance x1 of the plurality of instances”
“updating the first neural network at a first time based on the PTM to obtain a second neural network”
As drafted, are additional elements that amount to no more than mere instructions to apply an exception for the abstract ideas. See MPEP 2106.05(f).
The limitations:
“obtaining a data set, the data set comprising a plurality of instances (x1,....,xn) and a plurality of labels (p,, ...,y), each label of the plurality of labels corresponding to respective ones of the plurality of instances”
“outputting a plurality of corrected labels produced by the second neural network as a scrubbed data set”
As drafted, are additional elements that amount to no more than insignificant extra-solution activity. See MPEP 2106.05(g).
Therefore, the additional elements do not integrate the abstract ideas into a practical application.
Step 2B Analysis: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, all of the additional elements are “mere instructions to apply an exception” and “insignificant extra-solution activity”. Specifically, the outputting and obtaining a data set limitations recite the well-understood, routine, and conventional activity of receiving and transmitting data over a network. MPEP 2106.05(d)(II); OIP Techs., Inc v. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network). Mere instructions to apply an exception and insignificant extra-solution activity cannot provide an inventive concept. The claim is not patent eligible.
Regarding Claim 2,
Claim 2 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 2 is directed to a computer-implemented method of training a neural network, which is directed to a process, one of the statutory categories.
Step 2A Prong One Analysis: See corresponding analysis of claim 1.
Step 2A Prong Two Analysis: The judicial exceptions are not integrated into a practical application. In particular, the claim recited additional elements that are additional details that do not apply the exception in a meaningful way (See MPEP 2106.05(e)).
The limitations:
“wherein the second neural network is more robust to a label noise than the first neural network”
As drafted, are additional elements that do not apply an exception for the abstract ideas in a meaningful way. See MPEP 2106.05(e).
Therefore, the additional elements do not integrate the abstract ideas into a practical application.
Step 2B Analysis: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to the integration of the abstract ideas into a practical application, all of the additional elements do not apply the exception in a meaningful way. The claim is not patent eligible.
Regarding Claim 3,
Claim 3 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 3 is directed to a computer-implemented method of training a neural network, which is directed to a process, one of the statutory categories.
Step 2A Prong One Analysis: The limitations:
“estimating a corrected label ŷ1 of the first instance x1”
“forming a second data set from the data set by replacing ỹ1 with ŷ1”
“repeating the estimating and forming for remaining instances of the data set, wherein the plurality of corrected labels comprises ŷ1, ..., ŷn”
As drafted, under their broadest reasonable interpretations, cover mental processes,
i.e., concepts performed in the human mind (including an observation, evaluation,
judgement, opinion). The above limitations in the context of this claim correspond to
mental processes, e.g., evaluation and judgement with assistance of pen and paper.
Step 2A Prong Two Analysis: The judicial exceptions are not integrated into a practical application. In particular, the claim recites additional elements that are mere instructions to apply (See MPEP 2106.05(f)).
The limitations:
“using the second neural network”
As drafted, are additional elements that amount to no more than mere instructions to apply an exception for the abstract ideas. See MPEP 2106.05(f).
Therefore, the additional elements do not integrate the abstract ideas into a practical application.
Step 2B Analysis: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, all of the additional elements are “mere instructions to apply an exception” Mere instructions to apply an exception cannot provide an inventive concept. The claim is not patent eligible.
Regarding Claim 4,
Claim 4 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 4 is directed to a computer-implemented method of training a neural network, which is directed to a process, one of the statutory categories.
Step 2A Prong One Analysis: See corresponding analysis of claim 3.
Step 2A Prong Two Analysis: The judicial exceptions are not integrated into a practical application. In particular, the claim recites additional elements that are insignificant extra-solution activity (See MPEP 2106.05(g)).
The limitations:
“wherein the obtaining the data set comprises receiving the data set from a user”
“wherein the outputting the scrubbed data set comprises outputting the scrubbed data set to the user”
As drafted, are additional elements that amount to no more than insignificant extra-solution activity. See MPEP 2106.05(g).
Therefore, the additional elements do not integrate the abstract ideas into a practical application.
Step 2B Analysis: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, all of the additional elements are “insignificant extra-solution activity”. Specifically, the obtaining and outputting limitations recite the well-understood, routine, and conventional activity of receiving and transmitting data over a network. MPEP 2106.05(d)(II); OIP Techs., Inc v. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network). Insignificant extra-solution activity cannot provide an inventive concept. The claim is not patent eligible.
Regarding Claim 5,
Claim 5 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 5 is directed to a computer-implemented method of training a neural network, which is directed to a process, one of the statutory categories.
Step 2A Prong One Analysis: See corresponding analysis of claim 1.
Step 2A Prong Two Analysis: The judicial exceptions are not integrated into a practical application. In particular, the claim recited additional elements that are additional details that do not apply the exception in a meaningful way (See MPEP 2106.05(e)).
The limitations:
“wherein the data set includes first information indicating accidental ad clicks, second information indicating fraud clicks, and third information including delayed feedback”
As drafted, are additional elements that do not apply an exception for the abstract ideas in a meaningful way. See MPEP 2106.05(e).
Therefore, the additional elements do not integrate the abstract ideas into a practical application.
Step 2B Analysis: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to the integration of the abstract ideas into a practical application, all of the additional elements do not apply the exception in a meaningful way. The claim is not patent eligible.
Regarding Claim 6,
Claim 6 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 6 is directed to a computer-implemented method of training a neural network, which is directed to a process, one of the statutory categories.
Step 2A Prong One Analysis: See corresponding analysis of claim 1.
Step 2A Prong Two Analysis: The judicial exceptions are not integrated into a practical application. In particular, the claim recites additional elements that are insignificant extra-solution activity (See MPEP 2106.05(g)).
The limitations:
“wherein the obtaining the data set comprises receiving the data set from an ad exchange server”
As drafted, are additional elements that amount to no more than insignificant extra-solution activity. See MPEP 2106.05(g).
Therefore, the additional elements do not integrate the abstract ideas into a practical application.
Step 2B Analysis: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, all of the additional elements are “insignificant extra-solution activity”. Specifically, the obtaining/receiving limitation recites the well-understood, routine, and conventional activity of receiving and transmitting data over a network. MPEP 2106.05(d)(II); OIP Techs., Inc v. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network). Insignificant extra-solution activity cannot provide an inventive concept. The claim is not patent eligible.
Regarding Claim 7,
Claim 7 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 7 is directed to a computer-implemented method of training a neural network, which is directed to a process, one of the statutory categories.
Step 2A Prong One Analysis: The limitations:
“computing user response prediction (URP)”
As drafted, under their broadest reasonable interpretations, cover mental processes,
i.e., concepts performed in the human mind (including an observation, evaluation,
judgement, opinion). The above limitations in the context of this claim correspond to
mental processes, e.g., evaluation and judgement with assistance of pen and paper.
Step 2A Prong Two Analysis: The judicial exceptions are not integrated into a practical application. In particular, the claim recites additional elements that are mere instructions to apply (See MPEP 2106.05(f)) and insignificant extra-solution activity (See MPEP 2106.05(g)).
The limitations:
“by inputting the bid request to the second neural network”
As drafted, are additional elements that amount to no more than mere instructions to apply an exception for the abstract ideas. See MPEP 2106.05(f).
The limitations:
“receiving, from the ad exchange server, a bid request”
As drafted, are additional elements that amount to no more than insignificant extra-solution activity. See MPEP 2106.05(g).
Therefore, the additional elements do not integrate the abstract ideas into a practical application.
Step 2B Analysis: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, all of the additional elements are “mere instructions to apply an exception” and “insignificant extra-solution activity”. Specifically, the receiving limitation recites the well-understood, routine, and conventional activity of receiving and transmitting data over a network. MPEP 2106.05(d)(II); OIP Techs., Inc v. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network). Mere instructions to apply an exception and insignificant extra-solution activity cannot provide an inventive concept. The claim is not patent eligible.
Regarding Claim 8,
Claim 8 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 8 is directed to a computer-implemented method of training a neural network, which is directed to a process, one of the statutory categories.
Step 2A Prong One Analysis: The limitations:
“wherein the obtaining the first transition matrix (PTM) comprises minimizing a norm of the PTM”
As drafted, under their broadest reasonable interpretations, cover mathematical concepts, i.e., mathematical relationships, mathematical formulas or equations, and mathematical calculations. The above limitations in the context of this claim correspond to mathematical relationships and calculations.
Step 2A Prong Two Analysis: The judicial exceptions are not integrated into a practical application. In particular, the claim recites additional elements that are mere instructions to apply (See MPEP 2106.05(f)).
The limitations:
“after a warm-up training”
As drafted, are additional elements that amount to no more than mere instructions to apply an exception for the abstract ideas. See MPEP 2106.05(f).
Therefore, the additional elements do not integrate the abstract ideas into a practical application.
Step 2B Analysis: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, all of the additional elements are “mere instructions to apply an exception”. Mere instructions to apply an exception cannot provide an inventive concept. The claim is not patent eligible.
Regarding Claim 9,
Claim 9 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 9 is directed to a computer-implemented method of training a neural network, which is directed to a process, one of the statutory categories.
Step 2A Prong One Analysis: See corresponding analysis of claim 8.
Step 2A Prong Two Analysis: The judicial exceptions are not integrated into a practical application. In particular, the claim recited additional elements that are additional details that do not apply the exception in a meaningful way (See MPEP 2106.05(e)).
The limitations:
“wherein the norm is a Frobenius norm, and a solution to find the PTM with the Frobenius norm is based on noisy labels comprised in the data set and based on the PTM”
As drafted, are additional elements that do not apply an exception for the abstract ideas in a meaningful way. See MPEP 2106.05(e).
Therefore, the additional elements do not integrate the abstract ideas into a practical application.
Step 2B Analysis: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to the integration of the abstract ideas into a practical application, all of the additional elements do not apply the exception in a meaningful way. The claim is not patent eligible.
Regarding Claim 10,
Claim 10 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 10 is directed to a computer-implemented method of training a neural network, which is directed to a process, one of the statutory categories.
Step 2A Prong One Analysis: The limitations:
“minimizing the norm of the PTM comprises: initializing all entries of the PTM to 0”
“computing f for an index i = 1”
“obtaining a classification j = argmax(f(x1))”
“setting column j of the PTM equal to the corresponding entries of the output”
“repeating the obtaining and the setting for the index i = 2, ..., n”
As drafted, under their broadest reasonable interpretations, cover mathematical concepts, i.e., mathematical relationships, mathematical formulas or equations, and mathematical calculations. The above limitations in the context of this claim correspond to mathematical relationships and calculations.
Step 2A Prong Two Analysis: The judicial exceptions are not integrated into a practical application. In particular, the claim recited additional elements that are additional details that do not apply the exception in a meaningful way (See MPEP 2106.05(e)).
The limitations:
“wherein the norm is a Frobenius norm”
As drafted, are additional elements that do not apply an exception for the abstract ideas in a meaningful way. See MPEP 2106.05(e).
Therefore, the additional elements do not integrate the abstract ideas into a practical application.
Step 2B Analysis: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to the integration of the abstract ideas into a practical application, all of the additional elements do not apply the exception in a meaningful way. The claim is not patent eligible.
Regarding Claim 12,
Claim 12 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 12 is directed to a computer-implemented method of training a neural network, which is directed to a process, one of the statutory categories.
Step 2A Prong One Analysis: The limitations:
“the combining the PTM with the second transition matrix (NTM) to obtain a third transition matrix (WKM) is performed for the index i”
As drafted, under their broadest reasonable interpretations, cover mathematical concepts, i.e., mathematical relationships, mathematical formulas or equations, and mathematical calculations. The above limitations in the context of this claim correspond to mathematical relationships and calculations.
Step 2A Prong Two Analysis: The judicial exceptions are not integrated into a practical application. In particular, the claim recites additional elements that are mere instructions to apply (See MPEP 2106.05(f)).
The limitations:
“updating the neural network the first time based on the WKM is performed for the index i”
As drafted, are additional elements that amount to no more than mere instructions to apply an exception for the abstract ideas. See MPEP 2106.05(f).
Therefore, the additional elements do not integrate the abstract ideas into a practical application.
Step 2B Analysis: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, all of the additional elements are “mere instructions to apply an exception”. Mere instructions to apply an exception cannot provide an inventive concept. The claim is not patent eligible.
Regarding Claim 13,
Claim 13 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 13 is directed to an apparatus for training a neural network neural network, which is directed to a machine, one of the statutory categories.
Step 2A Prong One Analysis: The limitations:
“obtain a probability of a noisy label (P) given xi”
“obtain a first transition matrix (PTM), wherein the obtaining the PTM comprises including in the PTM a term based on the first output and P”
“combine the PTM with a second transition matrix (NTM) to obtain a third transition matrix (WKM)”
“wherein the updating comprises minimizing a loss function based on the WKM”
As drafted, under their broadest reasonable interpretations, cover mathematical concepts, i.e., mathematical relationships, mathematical formulas or equations, and mathematical calculations. The above limitations in the context of this claim correspond to mathematical relationships and calculations.
Step 2A Prong Two Analysis: The judicial exceptions are not integrated into a practical application. In particular, the claim recites additional elements that are mere instructions to apply (See MPEP 2106.05(f)) and insignificant extra-solution activity (See MPEP 2106.05(g)).
The limitations:
“An apparatus for training a neural network neural network, the apparatus comprising: one or more processors; and one or more memories, the one or more memories storing instructions”
“train the neural network using the data set to obtain a first neural network”
“obtain a first output of the first neural network in response to a first instance (xi) of the plurality of instances”
“update the first neural network at a first time based on the PTM to obtain a second neural network”
As drafted, are additional elements that amount to no more than mere instructions to apply an exception for the abstract ideas. See MPEP 2106.05(f).
The limitations:
“obtain a data set, the data set comprising a plurality of instances (x1,....,xn) and a plurality of labels (f,, ...,y), each label of the plurality of labels corresponding to respective ones of the plurality of instances”
“output a plurality of corrected labels produced by the second neural network as a scrubbed data set”
As drafted, are additional elements that amount to no more than insignificant extra-solution activity. See MPEP 2106.05(g).
Therefore, the additional elements do not integrate the abstract ideas into a practical application.
Step 2B Analysis: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, all of the additional elements are “mere instructions to apply an exception” and “insignificant extra-solution activity”. Specifically, the outputting and obtaining a data set limitations recite the well-understood, routine, and conventional activity of receiving and transmitting data over a network. MPEP 2106.05(d)(II); OIP Techs., Inc v. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network). Mere instructions to apply an exception and insignificant extra-solution activity cannot provide an inventive concept. The claim is not patent eligible.
Regarding Claim 14,
Claim 14 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 14 is directed to an apparatus for training a neural network neural network, which is directed to a machine, one of the statutory categories.
Step 2A Prong One Analysis: The limitations:
“estimate a corrected label ŷ1 of the first instance x1”
“form a second data set from the data set by replacing ỹ1 with ŷ1”
“repeat by repeating estimating and forming for remaining instances of the data set, wherein a plurality of corrected labels comprises ŷ1, ..., ŷn”
As drafted, under their broadest reasonable interpretations, cover mental processes,
i.e., concepts performed in the human mind (including an observation, evaluation,
judgement, opinion). The above limitations in the context of this claim correspond to
mental processes, e.g., evaluation and judgement with assistance of pen and paper.
Step 2A Prong Two Analysis: The judicial exceptions are not integrated into a practical application. In particular, the claim recites additional elements that are mere instructions to apply (See MPEP 2106.05(f)).
The limitations:
“using the second neural network”
As drafted, are additional elements that amount to no more than mere instructions to apply an exception for the abstract ideas. See MPEP 2106.05(f).
Therefore, the additional elements do not integrate the abstract ideas into a practical application.
Step 2B Analysis: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, all of the additional elements are “mere instructions to apply an exception”. Mere instructions to apply an exception cannot provide an inventive concept. The claim is not patent eligible.
Regarding Claim 15,
Claim 15 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 15 is directed to an apparatus for training a neural network neural network, which is directed to a machine, one of the statutory categories.
Step 2A Prong One Analysis: See corresponding analysis of claim 14.
Step 2A Prong Two Analysis: The judicial exceptions are not integrated into a practical application. In particular, the claim recites additional elements that are insignificant extra-solution activity (See MPEP 2106.05(g)).
The limitations:
“obtain the data set by receiving the data set from a user”
“output the scrubbed data set by outputting the scrubbed data set to the user”
As drafted, are additional elements that amount to no more than insignificant extra-solution activity. See MPEP 2106.05(g).
Therefore, the additional elements do not integrate the abstract ideas into a practical application.
Step 2B Analysis: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, all of the additional elements are “insignificant extra-solution activity”. Specifically, the obtaining and outputting limitations recite the well-understood, routine, and conventional activity of receiving and transmitting data over a network. MPEP 2106.05(d)(II); OIP Techs., Inc v. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network). Insignificant extra-solution activity cannot provide an inventive concept. The claim is not patent eligible.
Regarding Claim 16,
Claim 16 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 16 is directed to an apparatus for training a neural network neural network, which is directed to a machine, one of the statutory categories.
Step 2A Prong One Analysis: See corresponding analysis of claim 13.
Step 2A Prong Two Analysis: The judicial exceptions are not integrated into a practical application. In particular, the claim recites additional elements that are insignificant extra-solution activity (See MPEP 2106.05(g)).
The limitations:
“obtain the data set by receiving the data set from an ad exchange server”
As drafted, are additional elements that amount to no more than insignificant extra-solution activity. See MPEP 2106.05(g).
Therefore, the additional elements do not integrate the abstract ideas into a practical application.
Step 2B Analysis: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, all of the additional elements are “insignificant extra-solution activity”. Specifically, the obtaining/receiving limitation recites the well-understood, routine, and conventional activity of receiving and transmitting data over a network. MPEP 2106.05(d)(II); OIP Techs., Inc v. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network). Insignificant extra-solution activity cannot provide an inventive concept. The claim is not patent eligible.
Regarding Claim 17,
Claim 17 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 17 is directed to an apparatus for training a neural network neural network, which is directed to a machine, one of the statutory categories.
Step 2A Prong One Analysis: The limitations:
“compute a user response prediction (URP)”
As drafted, under their broadest reasonable interpretations, cover mental processes,
i.e., concepts performed in the human mind (including an observation, evaluation,
judgement, opinion). The above limitations in the context of this claim correspond to
mental processes, e.g., evaluation and judgement with assistance of pen and paper.
Step 2A Prong Two Analysis: The judicial exceptions are not integrated into a practical application. In particular, the claim recites additional elements that are mere instructions to apply (See MPEP 2106.05(f)) and insignificant extra-solution activity (See MPEP 2106.05(g)).
The limitations:
“by inputting the bid request to the second neural network”
As drafted, are additional elements that amount to no more than mere instructions to apply an exception for the abstract ideas. See MPEP 2106.05(f).
The limitations:
“receiving, from the ad exchange server, a bid request”
As drafted, are additional elements that amount to no more than insignificant extra-solution activity. See MPEP 2106.05(g).
Therefore, the additional elements do not integrate the abstract ideas into a practical application.
Step 2B Analysis: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, all of the additional elements are “mere instructions to apply an exception” and “insignificant extra-solution activity”. Specifically, the receiving limitation recites the well-understood, routine, and conventional activity of receiving and transmitting data over a network. MPEP 2106.05(d)(II); OIP Techs., Inc v. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network). Mere instructions to apply an exception and insignificant extra-solution activity cannot provide an inventive concept. The claim is not patent eligible.
Regarding Claim 18,
Claim 18 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 18 is directed to an apparatus for training a neural network neural network, which is directed to a machine, one of the statutory categories.
Step 2A Prong One Analysis: The limitations:
“obtain the first transition matrix (PTM) by minimizing a norm of the PTM”
As drafted, under their broadest reasonable interpretations, cover mathematical concepts, i.e., mathematical relationships, mathematical formulas or equations, and mathematical calculations. The above limitations in the context of this claim correspond to mathematical relationships and calculations.
Step 2A Prong Two Analysis: The judicial exceptions are not integrated into a practical application. In particular, the claim recites additional elements that are mere instructions to apply (See MPEP 2106.05(f)).
The limitations:
“after a warm-up training”
As drafted, are additional elements that amount to no more than mere instructions to apply an exception for the abstract ideas. See MPEP 2106.05(f).
Therefore, the additional elements do not integrate the abstract ideas into a practical application.
Step 2B Analysis: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, all of the additional elements are “mere instructions to apply an exception”. Mere instructions to apply an exception cannot provide an inventive concept. The claim is not patent eligible.
Regarding Claim 20,
Claim 20 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 20 is directed to a non-transitory computer readable medium storing instructions for training a neural network, which is directed to a machine, one of the statutory categories.
Step 2A Prong One Analysis: The limitations:
“obtain a probability of a noisy label (P) given xi”
“obtain a first transition matrix (PTM), wherein the obtaining the PTM comprises including in the PTM a term based on the first output and P”
“combine the PTM with a second transition matrix (NTM) to obtain a third transition matrix (WKM)”
“wherein the updating comprises minimizing a loss function based on the WKM”
As drafted, under their broadest reasonable interpretations, cover mathematical concepts, i.e., mathematical relationships, mathematical formulas or equations, and mathematical calculations. The above limitations in the context of this claim correspond to mathematical relationships and calculations.
Step 2A Prong Two Analysis: The judicial exceptions are not integrated into a practical application. In particular, the claim recites additional elements that are mere instructions to apply (See MPEP 2106.05(f)) and insignificant extra-solution activity (See MPEP 2106.05(g)).
The limitations:
“A non-transitory computer readable medium storing instructions for training a neural network, the instructions configured to cause a computer to at least…”
“train a neural network using the data set to obtain a first neural network”
“obtain a first output of the first neural network in response to a first instance xi of the plurality of instances”
“update the first neural network at a first time based on the PTM to obtain a second neural network”
As drafted, are additional elements that amount to no more than mere instructions to apply an exception for the abstract ideas. See MPEP 2106.05(f).
The limitations:
“obtain a data set, the data set comprising a plurality of instances (x1,....,xn) and a plurality of labels (p,, ...,y), each label of the plurality of labels corresponding to respective ones of the plurality of instances”
“output a plurality of corrected labels produced by the second neural network as a scrubbed data set”
As drafted, are additional elements that amount to no more than insignificant extra-solution activity. See MPEP 2106.05(g).
Therefore, the additional elements do not integrate the abstract ideas into a practical application.
Step 2B Analysis: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, all of the additional elements are “mere instructions to apply an exception” and “insignificant extra-solution activity”. Specifically, the outputting and obtaining a data set limitations recite the well-understood, routine, and conventional activity of receiving and transmitting data over a network. MPEP 2106.05(d)(II); OIP Techs., Inc v. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network). Mere instructions to apply an exception and insignificant extra-solution activity cannot provide an inventive concept. The claim is not patent eligible.
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.
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.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
Claims 1-4, 12-15, and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Levanony et al. (U.S. Patent Publication No. 2021/0406608) (“Levanony”) in view of Gutierrez et al. (U.S. Patent Publication No. 2022/0215141) (“Gutierrez”).
Regarding claim 1, Levanony teaches a computer-implemented method of training a neural network, the computer-implemented method comprising: obtaining a data set, the data set comprising a plurality of instances (x1, ... , xn) and a plurality of labels (ỹ1 , ... , ỹn), each label of the plurality of labels corresponding to respective ones of the plurality of instances (Levanony [0018] “In the example of FIG. 1, a trained classifier 106 may be used to generate annotated data 108 including labels and associated positive and negative error rates for each of the labels. For example, labeled data may be split into subsets including a training set (not shown) and a test set 104.” Levanony provides annotated data 108, which include labels, corresponding to obtaining a data set comprising corresponding data instances and labels.); training the neural network using the data set to obtain a first neural network (Levanony [0017] “FIG. 1 includes unlabeled data 102 and test set 104 of labeled data shown being received at a trained classifier 106. For example, the trained classifier may be a neural network.”; [0018] “The trained classifier 106 may be trained using the training set to annotate unlabeled data.” Levanony provides training classifier 106, which is a neural network, using the training set subset of data 108 corresponding to training the neural network using the data set to obtain a first neural network.); obtaining a first output of the first neural network in response to a first instance x1 of the plurality of instances (Levanony [0018] “The trained classifier 106 may then be used to generate positive and negative error rates for each of a number of labels based on the test set 104” Levanony provides obtaining an output of the trained classifier 106 in response to the test subset of data 108 corresponding to obtaining a first output of the first neural network in response to a first instance of the plurality of instances.); obtaining a probability of a noisy label (P) given x1 (Levanony [0018] “For example, using the labels of the test set 104, an evaluator (not shown) may compute the negative error rate indicating the probability of a label being flipped into not being labeled under noise, and a positive error rate indicating the probability of an object not labeled being labeled under noise.” Levanony provides a positive error rate indicating the probability of an object not labeled being labeled under noise corresponding to obtaining a probability of a noisy label (P) given x1.); obtaining a first transition matrix (PTM), wherein the obtaining the PTM comprises including in the PTM a term based on the first output and P (Levanony [0019] “The flipped labels calculator 110 receives the annotated data 108 with associated positive and negative error rates and generates a flipped labels probability matrix 112 for the labels based on the positive and negative error rates. For example, the error rates can be used as a probability that an image has flipped labels. In various examples, given a known probability of each class to have flipped labels, the flipped labels probability matrix 112 may be a noise transition matrix” Levanony provides a probability matrix, which may be a noise transition matrix, which includes the outputted probability corresponding to obtaining a first transition matrix (PTM), wherein the obtaining the PTM comprises including in the PTM a term based on the first output and P.); combining the PTM with a second transition matrix (NTM) to obtain a third transition matrix (WKM) (Levanony 0019] “In various examples, given a known probability of each class to have flipped labels, the flipped labels probability matrix 112 may be a noise transition matrix”; [0021] “Given a proper composite loss custom-character, a forward correction loss custom-character.sub.ψ.sup..fwdarw. may be defined by the Equation: Eq 1 where T.sup.T is the transpose of flipped labeled probability matrix T, ψ.sup.−1 is the inverse of link function ψ, and h(x) is a transformation function representing the transformations of the intermediate layers of a neural network.” Levanony provides matrix operations including utilizing a matrix transpose for a noise transition matrix, as shown in Equation 1 corresponding to combining the PTM with a second transition matrix (NTM) to obtain a third transition matrix (WKM).), updating the first neural network at a first time based on the PTM to obtain a second neural network (Levanony [0017] “The system 100 further includes a second classifier 114 communicatively coupled to the trained classifier 106 and the flipped labels calculator 110. For example, the second classifier 114 is shown being trained on the annotated data 108 using the flipped labels probability matrix 112.”; [0020] “The second classifier 114 is then trained using a loss based on the flipped labeled probability matrix 112 and the annotated data 108. The second classifier 114 may be trained to classify input annotated data 108 by treating labels as flipped according to the flipped labeled probability matrix 112. The second classifier 114 may be trained using a method that reverts the flipped labels to train such that the second classifier 114 outputs classifications as if the labels were not flipped. For example, the flipped labels may be reverted using a forward corrected loss or a backward corrected loss.” Levanony provides training a second classifier using the probability matrix, data 108, and loss values from the first classifier corresponding to updating the first neural network at a first time based on the PTM to obtain a second neural network.), wherein the updating comprises minimizing a loss function based on the WKM (Levanony [0021] “For example, a composite loss can be expressed by the aid of a link function, and a proper composite loss includes a minimizer that assumes a particular shape of the link function applied to the class-conditional probabilities p(y|x).”; [0020] “The second classifier 114 is then trained using a loss based on the flipped labeled probability matrix 112 and the annotated data 108. The second classifier 114 may be trained to classify input annotated data 108 by treating labels as flipped according to the flipped labeled probability matrix 112. The second classifier 114 may be trained using a method that reverts the flipped labels to train such that the second classifier 114 outputs classifications as if the labels were not flipped. For example, the flipped labels may be reverted using a forward corrected loss or a backward corrected loss.” Levanony provides loss minimization for producing an updated neural network.).
Levanony fails to explicitly teach outputting a plurality of corrected labels produced by the second neural network as a scrubbed data set.
However, Gutierrez teaches outputting a plurality of corrected labels produced by the second neural network as a scrubbed data set (Gutierrez [0057] “A generative model, as used herein, is used to describe models that generate instances of output variables that may be used for machine learning…A generative adversarial network (GAN) is generally referred to as a machine learning framework in which two neural networks compete against each other (e.g., based on game theory).”; [0144] “A previously-trained machine learning model may be obtained in step 1301… For example, a previously-trained machine learning model may, in step 1301, may read and suggest labels for the fields of the retrieved dataset (retrieved in step 1300).”; [0145] “Once the data fields have been labeled, the true-source data may be scrubbed in step 1302, to selectively replace the content of fields based on the labels of the fields… Additionally or alternatively, in step 1302, users may be permitted to manually set data types and/or scrubbing policy. In step 1303, a scrubbed dataset may be generated.” Gutierrez provides a previously-trained machine learning model including a generative network consisting of two neural networks, which corresponds to the corrected labels produced by the second neural network, wherein a scrubbed dataset is subsequently generated once the corrected labels are provided, corresponding to outputting a plurality of corrected labels produced by the second neural network as a scrubbed data set.).
Levanony and Gutierrez are both considered to be analogous to the claimed invention because they are in the same field of artificial intelligence and more specifically applied to data labeling. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Levanony with the above teachings of Gutierrez. Doing so would allow for creation of a dataset to improve the accuracy of output (Gutierrez [0158] “Based on those dependencies, the relationships between the fields may be mapped within the dataset to improve the accuracy of the output. The output of this process is a generative machine learning model that is tied to the scrubbed true-source dataset and may be used in the subsequent step to generate synthetic data that follows the distributions of that scrubbed true-source data.”).
Regarding claim 2, Levanony in view of Gutierrez teaches the computer-implemented method of claim 1 as discussed above in the rejection of claim 1, wherein the second neural network is more robust to a label noise than the first neural network (Levanony [0064] “For example, the trainer module 426 can train the first classifier module 432 to classify a training set of the labeled data. In some examples, the trainer module 426 can train a second classifier using the annotated data and associated error rates. For example, an embodiment in which the trainer module 426 trains the second classifier using a forward corrected loss has the advantage of noise robustness. In some examples, an embodiment in which the trainer module 426 trains the second classifier using a backward corrected loss has the advantage of being differentiable and being able to be minimized with any suitable back-propagation algorithm” Levanony provides training a second classifier based on a loss from a first classifier, which has the advantage of noise robustness, corresponding to the second neural network is more robust to a label noise than the first neural network.).
It would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Levanony in view of Gutierrez for the same reasons disclosed above in the rejection of claim 1.
Regarding claim 3, Levanony in view of Gutierrez teaches the computer-implemented method of claim 1, as discussed above in the rejection of claim 1, further comprising: estimating a corrected label ŷ1 of the first instance x1 using the second neural network (Gutierrez [0144] “For example, a previously-trained machine learning model may, in step 1301, may read and suggest labels for the fields of the retrieved dataset (retrieved in step 1300).” Gutierrez provides machine learning models for suggesting labels for a retrieved dataset, corresponding to estimating a corrected label ŷ1 of the first instance x1 using the second neural network.); forming a second data set from the data set by replacing ỹ1 with ŷ1 (Gutierrez [0145] “Once the data fields have been labeled, the true-source data may be scrubbed in step 1302, to selectively replace the content of fields based on the labels of the fields. For instance, fields having been labeled with labels identifying sensitive information (e.g., names, addresses, account numbers, etc.) may be replaced with a contextually similar alternative value that follows the same schema as the source field.” Gutierrez provides replacing data in a dataset based on label information corresponding to forming a second data set from the data set by replacing ỹ1 with ŷ1.); and repeating the estimating and forming for remaining instances of the data set, wherein the plurality of corrected labels comprises ŷ1, ..., ŷn (Gutierrez [0145] “Once the data fields have been labeled, the true-source data may be scrubbed in step 1302, to selectively replace the content of fields based on the labels of the fields. For instance, fields having been labeled with labels identifying sensitive information (e.g., names, addresses, account numbers, etc.) may be replaced with a contextually similar alternative value that follows the same schema as the source field. The replacement technique may be the same for all fields having been labeled with a label identifying the field as containing sensitive information.” Gutierrez provides suggesting labels and replacing data for data in a dataset that is deemed as identifying sensitive information corresponding to repeating the estimating and forming for remaining instances of the data set, wherein a plurality of corrected labels comprises ŷ1, ..., ŷn.).
Levanony and Gutierrez are both considered to be analogous to the claimed invention because they are in the same field of artificial intelligence and more specifically applied to data labeling. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Levanony with the above teachings of Gutierrez. Doing so would allow for creation of a dataset to improve the accuracy of output (Gutierrez [0158] “Based on those dependencies, the relationships between the fields may be mapped within the dataset to improve the accuracy of the output. The output of this process is a generative machine learning model that is tied to the scrubbed true-source dataset and may be used in the subsequent step to generate synthetic data that follows the distributions of that scrubbed true-source data.”).
Regarding claim 4, Levanony in view of Gutierrez teaches the computer-implemented method of claim 3 as discussed above in the rejection of claim 3, wherein the obtaining the data set comprises receiving the data set from a user (Gutierrez [0143] “FIG. 13 depicts a flow chart for a method of training a model based on true-source data. In step 1300, a true-source dataset is received. The starting of the process may be based on a request from a user or system desiring a synthetic dataset. The true-source dataset may be uploaded by the user or system or may be obtained from a remote user or system.” Gutierrez provides receiving the data set from a user.), and wherein the outputting the scrubbed data set comprises outputting the scrubbed data set to the user (Gutierrez [0145] “Additionally or alternatively, in step 1302, users may be permitted to manually set data types and/or scrubbing policy. In step 1303, a scrubbed dataset may be generated.” Gutierrez provides generating scrubbed data sets for users, wherein the policy may be set by a user, corresponding to outputting the scrubbed data set to the user.).
Levanony and Gutierrez are both considered to be analogous to the claimed invention because they are in the same field of artificial intelligence and more specifically applied to data labeling. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Levanony with the above teachings of Gutierrez. Doing so would allow for creation of a dataset to improve the accuracy of output (Gutierrez [0158] “Based on those dependencies, the relationships between the fields may be mapped within the dataset to improve the accuracy of the output. The output of this process is a generative machine learning model that is tied to the scrubbed true-source dataset and may be used in the subsequent step to generate synthetic data that follows the distributions of that scrubbed true-source data.”).
Regarding claim 12, Levanony in view of Gutierrez teaches the computer-implemented method of claim 1, as discussed above in the rejection of claim 1, wherein: the combining the PTM with the second transition matrix (NTM) to obtain a third transition matrix (WKM) is performed for the index I (Levanony [0019] “In various examples, given a known probability of each class to have flipped labels, the flipped labels probability matrix 112 may be a noise transition matrix”; [0021] “Given a proper composite loss custom-character, a forward correction loss custom-character.sub.ψ.sup..fwdarw. may be defined by the Equation: Eq 1 where T.sup.T is the transpose of flipped labeled probability matrix T, ψ.sup.−1 is the inverse of link function ψ, and h(x) is a transformation function representing the transformations of the intermediate layers of a neural network. ” Levanony provides matrix operations including utilizing a matrix transpose for a noise transition matrix, as shown in Equation 1 corresponding to combining the PTM with the second transition matrix (NTM) to obtain a third transition matrix (WKM) for a first index corresponding to the combining the PTM with the second transition matrix (NTM) to obtain a third transition matrix (WKM) is performed for the index I); the updating the neural network the first time based on the WKM is performed for the index i (Levanony [0021] “For example, a composite loss can be expressed by the aid of a link function, and a proper composite loss includes a minimizer that assumes a particular shape of the link function applied to the class-conditional probabilities p(y|x).”; [0020] “The second classifier 114 is then trained using a loss based on the flipped labeled probability matrix 112 and the annotated data 108. The second classifier 114 may be trained to classify input annotated data 108 by treating labels as flipped according to the flipped labeled probability matrix 112. The second classifier 114 may be trained using a method that reverts the flipped labels to train such that the second classifier 114 outputs classifications as if the labels were not flipped. For example, the flipped labels may be reverted using a forward corrected loss or a backward corrected loss.” Levanony provides loss minimization for producing an updated neural network for a first index corresponding to updating the neural network the first time based on the WKM is performed for the index i.).
It would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Levanony in view of Gutierrez for the same reasons disclosed above in the rejection of claim 1.
Regarding claim 13, it is the apparatus embodiment of claim 1 with similar limitations to claim 1 and is rejected using the same reasoning found above in the rejection of claim 1. Further, Levanony teaches an apparatus comprising: one or more processors (Levanony [0016] “According to embodiments of the present disclosure, system includes a processor that can evaluate a trained first classifier on a test set of labeled data to generate error rates for a number of labels” Levanony provides a processor.); and one or more memories, the one or more memories storing instructions (Levanony [0061] “The computing device 400 may include a processor 402 that is to execute stored instructions, a memory device 404 to provide temporary memory space for operations of said instructions during operation.” Levanony provides one or more memories, the one or more memories storing instructions.).
It would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Levanony in view of Gutierrez for the same reasons disclosed above in the rejection of claim 1.
Regarding claim 14, the rejection of claim 13 is incorporated herein. Further, the limitations in this claim are taught by Levanony in view of Gutierrez for the same reasons disclosed above in the rejection of claim 3.
Regarding claim 15, the rejection of claim 14 is incorporated herein. Further, the limitations in this claim are taught by Levanony in view of Gutierrez for the same reasons disclosed above in the rejection of claim 4.
Regarding claim 20, it is the non-transitory computer readable medium storing instructions embodiment of claim 1 with similar limitations to claim 1 and is rejected using the same reasoning found above in the rejection of claim 1. Further Levanony teaches a non-transitory computer readable medium storing instructions for training a neural network (Levanony [0078] “Referring now to FIG. 7, a block diagram is depicted of an example tangible, non-transitory computer-readable medium 700 that can train classifiers to classify data using unlabeled data. The tangible, non-transitory, computer-readable medium 700 may be accessed by a processor 702 over a computer interconnect 704. Furthermore, the tangible, non-transitory, computer-readable medium 700 may include code to direct the processor 702 to perform the operations of the methods 200A, 200B, and 300 of FIGS. 2A, 2B, and 3.” Levanony provides a non-transitory computer readable medium storing instructions.).
It would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Levanony in view of Gutierrez for the same reasons disclosed above in the rejection of claim 1.
Claim 5 is rejected under 35 U.S.C. 103 as being unpatentable over Levanony et al. (U.S. Patent Publication No. 2021/0406608) (“Levanony”) in view of Gutierrez et al. (U.S. Patent Publication No. 2022/0215141) (“Gutierrez”) in further view of Kudlacz (U.S. Patent Publication No. 2022/0180393) (“Kudlacz”).
Regarding claim 5, Levanony in view of Gutierrez teaches the computer-implemented method of claim 1 as discussed above in the rejection of claim 1, but fails to teach wherein the data set includes first information indicating accidental ad clicks, second information indicating fraud clicks, and third information including delayed feedback.
However, Kudlacz teaches wherein the data set includes first information indicating accidental ad clicks, second information indicating fraud clicks (Kudlacz [0044] “Activity related to served advertisement objects may include invalid activity. In general, in the context of the present disclosure, any interactions with served advertisement objects that are not by a real user having interest in the advertised product may be considered invalid activity. Invalid activity may include activity that is intentionally fraudulent or malicious (e.g., in the case of click fraud), as well as activity that is unintentionally invalid (e.g., a real interested user accidentally clicks on an advertisement twice).” Kudlacz provides obtaining data related to accidental and fraudulent advertisement clicks.), and third information including delayed feedback (Kudlacz [0057] “It should be appreciated that the operations performed by the advertising platform 100 to serve advertisement objects in response to requests (e.g., search queries) and to detect invalid activity should be performed in real-time or near real-time. For example, there should be negligible delay (e.g., a time delay of no more than a few 100 ms) between receiving a request and serving advertisement objects, so that a response to the request (e.g., displaying search results in response to a search query) is not delayed by the advertising platform 100.” Kudlacz provides “near real-time” invalid activity detection, corresponding to delayed feedback.).
Levanony, Gutierrez and Kudlacz are all considered to be analogous to the claimed invention because they are in the same field of artificial intelligence and more specifically user data collection and labeling. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Levanony in view of Gutierrez with the above teachings of Kudlacz. Doing so would allow for detecting and mitigating invalid activity related to served advertisement objects (Kudlacz [0047] “The advertising platform 100 includes an invalid activity detection module 160 to detect and mitigate invalid activity related to served advertisement objects.”).
Claim 6-7 and 16-17 are rejected under 35 U.S.C. 103 as being unpatentable over Levanony et al. (U.S. Patent Publication No. 2021/0406608) (“Levanony”) in view of Gutierrez et al. (U.S. Patent Publication No. 2022/0215141) (“Gutierrez”) in further view of Raush et al. (U.S. Patent Publication No. 2023/0214876) (“Raush”).
Regarding claim 6, Levanony in view of Gutierrez teaches the computer-implemented method of claim 1 as discussed above in the rejection of claim 1, but fails to teach wherein the obtaining the data set comprises receiving the data set from an ad exchange server.
However, Raush teaches wherein the obtaining the data set comprises receiving the data set from an ad exchange server (Raush [0029] “The user response prediction module 118 can receive a request (e.g., a bid request or the like as part of an auction process or the like) from a content presentation exchange 126 (e.g., an ad exchange or the like) associated with a client device of a user (e.g., via network 110). The user response prediction module 118 can extract a context (i.e., a plurality of categorical features) from or otherwise associated with the request to generate a feature vector x of predetermined size (e.g., containing 5, 10, 15 or other suitable number of categorical features). The information can be obtained from the request itself in addition to or alternatively to information that can be retrieved or otherwise derived from either or both of the user response data database 122 and/or the client device data database 124.” Raush provides obtaining context from an ad exchange server and generating feature vectors therefrom, corresponding to receiving the data set from an ad exchange server.).
Levanony, Gutierrez and Raush are all considered to be analogous to the claimed invention because they are in the same field of artificial intelligence and more specifically user data collection and labeling. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Levanony in view of Gutierrez with the above teachings of Raush. Doing so would allow for the attainment of features characterizing users and/or their device for the display of appropriate content or advertisements (Raush [0032] “These features may include at least one feature characterizing the user and at least one additional feature characterizing the client device. As stated above, the plurality of features may include information, such as, for example, data characterizing a publisher, a location of placement of the content presentation on a screen of the client device, data associated with the user such as, e.g., age, gender, and so forth. Further, the plurality of features may also include a geographic location specific to the user and/or the client device. The at least one feature characterizing the user may correspond to information such as age, gender, and so forth, and the at least one feature characterizing the client device may include a device identifier or other comparable information that characterizes the client device.”).
Regarding claim 7, Levanony in view Gutierrez in further view of Raush teaches the computer-implemented method of claim 6 as discussed above in the rejection of claim 6, further comprising: receiving, from the ad exchange server, a bid request (Raush [0029] “The user response prediction module 118 can receive a request (e.g., a bid request or the like as part of an auction process or the like) from a content presentation exchange 126 (e.g., an ad exchange or the like) associated with a client device of a user (e.g., via network 110).” Raush provides receiving, from the ad exchange server, a bid request.); and computing user response prediction (URP) by inputting the bid request to the second neural network (Raush [0029] “The user response prediction module 118 can receive a request (e.g., a bid request or the like as part of an auction process or the like) from a content presentation exchange 126 (e.g., an ad exchange or the like) associated with a client device of a user (e.g., via network 110).”; [0024] “Additionally or alternatively, the predictive model can be or include a classifier such as, for example, one or more linear classifiers (e.g., Fisher's linear discriminant, logistic regression, Naive Bayes classifier, and/or perceptron), support vector machines (e.g., least squares support vector machines), quadratic classifiers, kernel estimation models (e.g., k-nearest neighbor), boosting (meta-algorithm) models, decision trees (e.g., random forests, Gradient Boosting Trees), neural networks, and/or learning vector quantization models, etc.” Raush provides computing user response prediction with user response prediction module 118 which is a neural network that receives bid requests as input, corresponding to computing user response prediction (URP) by inputting the bid request to the second neural network.).
Levanony, Gutierrez and Raush are all considered to be analogous to the claimed invention because they are in the same field of artificial intelligence and more specifically user data collection and labeling. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Levanony in view of Gutierrez in further view of Raush. Doing so would allow for characterizing a likelihood of the user interacting with a content presentation (Raush [0004] “A predicted response value of the user to a content presentation can be generated using a predictive model and the feature vector. The predicted response value can characterize a likelihood of the user interacting with the content presentation.”).
Regarding claim 16, the rejection of claim 13 is incorporated herein. Further, the limitations in this claim are taught by Levanony in view of Gutierrez in further view of Raush for the same reasons disclosed above in the rejection of claim 6.
Regarding claim 17, the rejection of claim 16 is incorporated herein. Further, the limitations in this claim are taught by Levanony in view of Gutierrez in further view of Raush for the same reasons disclosed above in the rejection of claim 7.
Claim 8-10 and 18 are rejected under 35 U.S.C. 103 as being unpatentable over Levanony et al. (U.S. Patent Publication No. 2021/0406608) (“Levanony”) in view of Gutierrez et al. (U.S. Patent Publication No. 2022/0215141) (“Gutierrez”) in further view of Wu et al. (Multi-Class Classification from Noisy-Similarity-Labeled Data) (“Wu”).
Regarding claim 8, Levanony in view of Gutierrez teaches the computer-implemented method of claim 1 as discussed above in the rejection of claim 1, but fails to teach wherein the obtaining the first transition matrix (PTM) comprises minimizing a norm of the PTM after a warm-up training.
However, Wu teaches wherein the obtaining the first transition matrix (PTM) comprises minimizing a norm of the PTM after a warm-up training (Wu Section 4 Generalization error Theorem 1 “Assume the parameter matrices W1, . . . , Wd have Frobenius norm at most M1, . . . , Md, and the activation functions are 1-Lipschitz, positive-homogeneous, and applied element-wise (such as the ReLU). Assume the transition matrix is given, and the instances are upper bounded by B, i.e., kXk ≤ B for all X, and the loss function `(S ˆ¯ ii0, S¯ ii0) is upper bounded by M2 . en, for any δ > 0, with probability at least 1 − δ,”; Estimate noise transition matrix T Algorithm 1 “Stages 1-2: Learn Tˆ 1: Learn f(X) = Pˆ(Y¯ |X) by training the network in Figure 2 without the noise transition matrix layer; Minimize Lmns to learn g and stop when Pˆ(Y¯ |X) corresponds the minimum classification error on the noisy validation set” Wu provides a Frobenius norm and minimization for a noise transition matrix and training without the noise transition matrix layer, corresponding to obtaining the first transition matrix (PTM) comprises minimizing a norm of the PTM after a warm-up training.).
Levanony, Gutierrez and Wu are all considered to be analogous to the claimed invention because they are in the same field of artificial intelligence and more specifically noisy labels. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Levanony in view of Gutierrez with the above teachings of Wu. Doing so would allow for a learning system to learn a classifier which can assign noise-free class labels for instances (Wu Abstract “We further estimate the transition matrix from only noisy data and build a novel learning system to learn a classifier which can assign noise free class labels for instances”).
Regarding claim 9, Levanony in view of Gutierrez in further view of Wu teaches the computer-implemented method of claim 8 as discussed above in the rejection of claim 8, wherein the norm is a Frobenius norm, and a solution to find the PTM with the Frobenius norm is based on noisy labels comprised in the data set and based on the PTM (Wu Section 4 Generalization error Theorem 1 “Assume the parameter matrices W1, . . . , Wd have Frobenius norm at most M1, . . . , Md, and the activation functions are 1-Lipschitz, positive-homogeneous, and applied element-wise (such as the ReLU). Assume the transition matrix is given, and the instances are upper bounded by B, i.e., kXk ≤ B for all X, and the loss function `(S ˆ¯ ii0, S¯ ii0) is upper bounded by M2 . en, for any δ > 0, with probability at least 1 − δ,” Section 3.3 Estimate noise transition matrix T “However, the transition matrix is unknown. We will discuss how to estimate the transition matrix for the noisy-similarity-labeled data in this subsection.” Wu provides a solution to find the PTM with the Frobenius norm is based on noisy labels comprised in the data set and based on the PTM.).
Levanony, Gutierrez and Wu are all considered to be analogous to the claimed invention because they are in the same field of artificial intelligence and more specifically noisy labels. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Levanony in view of Gutierrez in further view of Wu with the above teachings of Wu. Doing so would allow for a learning system to learn a classifier which can assign noise-free class labels for instances (Wu Abstract “We further estimate the transition matrix from only noisy data and build a novel learning system to learn a classifier which can assign noise free class labels for instances”).
Regarding claim 10, Levanony in view of Gutierrez in further view of Wu teaches the computer-implemented method of claim 8 as discussed above in the rejection of claim 8, wherein the norm is a Frobenius norm (Wu Section 4 Generalization error Theorem 1 “Assume the parameter matrices W1, . . . , Wd have Frobenius norm at most M1” Wu provides a Frobenius norm.), and the minimizing the norm of the PTM comprises: initializing all entries of the PTM to 0 (Wu Section 3.3 Estimate noise transition matrix T “We illustrate that they can also be used to estimate the transition matrix for the noisy-similarity labeled data. Specifically, an anchor point x for class y is defined as P(Y = y|X = x) = 1 and P(Y = y 0 |X = x) = 0, ∀y 0 ∈ Y \ {y}. Let x be an anchor point for class i such that P(Y = i|X = x) = 1 and for k 6= i, P(Y = k|X = x) = 0.” Wu provides anchor points for conditional transition matrix initialization wherein all values may be initialized to 0 meeting certain conditions corresponding to initializing all entries of the PTM to 0.); computing f for an index i = 1; obtaining a classification j = argmax(f(x1)) (Wu Section 4 Generalization error “Then the output of the softmax function is defined as gi(X) = exp (hi(X))/ PC j=1 exp (hj (X)), i = 1, . . . , C, and f(X) = T >g(X) is the output of the noise transition matrix layer. Let ˆf = argmaxi∈{1,...,C} ˆfi be the classifier learned from the hypothesis space F determined by the neural network” Wu provides computing function outputs for a plurality of indices including a first index and obtaining a classification from a classifier using an argmax function of the computed function output.); setting column j of the PTM equal to the corresponding entries of the output; and repeating the obtaining and the setting for the index i = 2, ..., n (Wu Section 3.2 Likelihood-based estimator “For clarity, we visualize the predicted noisy similarity in Figure 3. If Xi and Xi 0 are predicted belonging to the same class, i.e., argmaxm∈C fm(Xi ; θ) = argmaxn∈C fn(Xi 0; θ), the predicted noisy similarity should be relatively high (S ˆ¯ ii0 = 0.30 in Figure 3(a)). By contrast, if Xi and Xi 0 are predicted belonging to dierent classes, the predicted noisy similarity should be relatively low (S ˆ¯ ii0 = 0.0654 in Figure 3(b)).” Wu provides calculating an argmax for different instances corresponding to setting column j of the PTM equal to the corresponding entries of the output; and repeating the obtaining and the setting for the index i = 2, ..., n.).
Levanony, Gutierrez and Wu are all considered to be analogous to the claimed invention because they are in the same field of artificial intelligence and more specifically noisy labels. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Levanony in view of Gutierrez in further view of Wu with the above teachings of Wu. Doing so would allow for a learning system to learn a classifier which can assign noise-free class labels for instances (Wu Abstract “We further estimate the transition matrix from only noisy data and build a novel learning system to learn a classifier which can assign noise free class labels for instances”).
Regarding claim 18, the rejection of claim 13 is incorporated herein. Further, the limitations in this claim are taught by Levanony in view of Gutierrez in further view of Wu for the same reasons disclosed above in the rejection of claim 8.
Response to Arguments
Applicant’s amendments overcome the previously applied objection to the specification and 35 U.S.C. 112 rejection.
Regarding the rejection applied under 35 U.S.C. 101, Applicant firstly asserts that the Present Application addresses a problem of a set of images being annotated with labels, but some of the labels are in error (“Remarks”, Page 11). Applicant further asserts that the Present Application solves the problem by estimating a noise transition matrix and a posterior clean label transition matrix (“Remarks”, Page 12). Applicant further asserts that a blending of the noise transition matrix and the posterior clean label transition matrix is used to obtain a posterior loss, which is used to train a classifier so that an improved classification vector is obtained. Applicant further asserts that a scrubbed dataset is obtained therefrom, with fewer label errors, which is an improvement and is obtained by the improved classifier using the blending of the matrices (“Remarks”, Page 12). Applicant therefore asserts that the claims provide a technical improvement (“Remarks”, Page 13).
However, even if the claims do recite an improvement in labeling accuracy, as written, it would be an improvement in the abstract idea of combining the PTM with a second transition matrix (NTM) to obtain a third transition matrix (WKM) and minimizing a loss function based on the WKM. As discussed in Applicant’s Remarks, the loss function is used to train the classifier so that an improved classification vector is obtained, thus providing the improvement. The MPEP notes that it is important to keep in mind that an improvement in the abstract idea itself is not an improvement in technology. MPEP 2106.05(a)(II).
Regarding the rejection applied under 35 U.S.C. 102/103, Applicant asserts that there is insufficient motivation to combine Levanony and Gutierrez and that the motivation statement does not have a rational underpinning and the 103 rejection is not prima facie, because Levanony does not have incorrectly labeled data and does not process noisy labels (“Remarks”, Page 15).
However, there is a motivation to combine the references. As discussed in Gutierrez [0158], the disclosed data labeling embodiments provide improved accuracy of the output for the generation of scrubbed datasets. Further Levanony and Gutierrez are analogous art, because they are in the same field of artificial intelligence and more specifically applied to data labeling. Further, Levanony does disclose “noisy labels” in [0021] and [0022] and calculating loss values for incorrectly labeled data. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Levanony with the above teachings of Gutierrez. Doing so would allow for creation of a dataset to improve the accuracy of output (Gutierrez [0158] “Based on those dependencies, the relationships between the fields may be mapped within the dataset to improve the accuracy of the output. The output of this process is a generative machine learning model that is tied to the scrubbed true-source dataset and may be used in the subsequent step to generate synthetic data that follows the distributions of that scrubbed true-source data.”).
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 KURT NICHOLAS PRESSLY whose telephone number is (703)756-4639. The examiner can normally be reached M-F 8-4.
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, Kamran Afshar can be reached at (571) 272-7796. 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.
/KURT NICHOLAS PRESSLY/Examiner, Art Unit 2125
/KAMRAN AFSHAR/Supervisory Patent Examiner, Art Unit 2125