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 .
Status of Claims
Claims 1 – 20 are pending and examined herein.
Claims 2-3, 10-11, 18-19 are rejected under 35 U.S.C. 112(b).
Claims 1 – 20 are rejected under 35 U.S.C. 101.
Claims 1 – 20 are rejected under 35 U.S.C. 103.
Specification
Applicant is reminded of the proper content of an abstract of the disclosure.
A patent abstract is a concise statement of the technical disclosure of the patent and should include that which is new in the art to which the invention pertains. The abstract should not refer to purported merits or speculative applications of the invention and should not compare the invention with the prior art.
If the patent is of a basic nature, the entire technical disclosure may be new in the art, and the abstract should be directed to the entire disclosure. If the patent is in the nature of an improvement in an old apparatus, process, product, or composition, the abstract should include the technical disclosure of the improvement. The abstract should also mention by way of example any preferred modifications or alternatives.
Where applicable, the abstract should include the following: (1) if a machine or apparatus, its organization and operation; (2) if an article, its method of making; (3) if a chemical compound, its identity and use; (4) if a mixture, its ingredients; (5) if a process, the steps.
Extensive mechanical and design details of an apparatus should not be included in the abstract. The abstract should be in narrative form and generally limited to a single paragraph within the range of 50 to 150 words in length.
See MPEP § 608.01(b) for guidelines for the preparation of patent abstracts.
The abstract of the disclosure is objected to because it is 157 words in length and is not concise as the disclosure permits. Also, the final sentence recites that “responsive message can be transmitted to a remote computing device including the risk indicator and the explanatory data.” With where including is located in the sentence, it is unclear whether that phrase modifies the responsive message or the remote computing device. A corrected abstract of the disclosure is required and must be presented on a separate sheet, apart from any other text. See MPEP § 608.01(b).
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.
Claims 2-3, 10-11, 18-19 are 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.
Claims 2, 10, and 18 recite “selecting a subset of parameter wavelet coefficients from the set of parameter wavelet coefficients that have parameter wavelet coefficients higher than remaining parameter wavelet coefficients in the set.” It is unclear what is meant by a parameter wavelet coefficient “having” another parameter wavelet coefficient because the parameter wavelet coefficients are themselves the members being selected. Therefore, it is unclear which coefficients are being compared. Also, it is unclear whether “higher” refers to a greater signed numerical value or a greater magnitude. In paragraph [0069] of the specification, it describes the basis functions as being ranked according to the magnitude of the corresponding parameter wavelet coefficients. The scope of the claimed selection cannot be determined with reasonable certainty. For examination purposes, the limitation would be read like “selecting, from the set of parameter wavelet coefficients, a subset comprising parameter wavelet coefficients having magnitudes greater than magnitudes of the remaining parameter wavelet coefficients”.
Claims 3, 11, 19 are dependent on claims 2, 10, 18. They do not resolve the issue of indefiniteness and are rejected with the same rationale.
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 - 20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
MPEP § 2109(III) sets out steps for evaluating whether a claim is drawn to patent-eligible subject matter. The analysis of claims 1 – 20, in accordance with these steps, follows.
Step 1 Analysis:
Step 1 is to determine whether the claim is directed to a statutory category (process, machine, manufacture, or composition of matter.
Claims 1 – 8 are directed to a method, meaning that it is directed to the statutory category of process. Claims 9 – 16 are directed to a system, which is the statutory category of machine. Claims 17 – 20 are directed to a non-transitory computer-readable medium comprising instruction, which can be an article of manufacture.
Step 2A Prong One, Step 2A Prong Two, and Step 2B Analysis:
Step 2A Prong One asks if the claim recites a judicial exception (abstract idea, law of nature, or natural phenomenon). If the claim recites a judicial exception, analysis proceeds to Step 2A Prong Two, which asks if the claim recites additional elements that integrate the abstract idea into a practical application. If the claim does not integrate the judicial exception, analysis proceeds to Step 2B, which asks if the claim amounts to significantly more than the judicial exception. If the claim does not amount to significantly more than the judicial exception, the claim is not eligible subject matter under 35 U.S.C. 101.
Regarding claim 1, the following claim elements are abstract ideas:
applying a risk prediction model to time-series data for an attribute associated with a target entity to generate a risk indicator for the target entity, (It broadly recites evaluating information concerning an entity and predicting the entity’s risk, which is practical to perform in the human mind under its broadest reasonable interpretation aside from the recitation of generic computer components or by a human using a pen and paper.)
applying a plurality of basis functions of a wavelet transformation on the parameters of the feature learning model to generate a set of parameter wavelet coefficients; (Applying basis functions of a wavelet transformation on the parameters is merely mathematical calculation, which is mathematical concept.)
generating explanatory data for the risk indicator based on the set of parameter wavelet coefficients; (It broadly recites evaluating the numerical coefficients to explain the risk indicator, which is practical to perform in the human mind under its broadest reasonable interpretation aside from the recitation of generic computer components or by a human using a pen and paper.)
The following claim elements are additional elements which, taken alone or in combination with the other additional elements, do not integrate the judicial exception into a practical application nor amount to significantly more than the judicial exception:
one or more processing devices performing operations comprising (This falls under mere instructions to apply an exception. See MPEP § 2106.05(f). Therefore, this does not amount to significantly more than the judicial exception.)
wherein the risk prediction model comprises a feature learning model configured to receive the time-series data as input and a risk classification model configured to receive output of the feature learning model and generate the risk indicator as output; (This falls under mere instructions to apply an exception. See MPEP § 2106.05(f). Therefore, this does not amount to significantly more than the judicial exception.)
accessing parameters of the feature learning model; (This is mere data gathering, an insignificant extra solution activity, which is a well-understood, routine conventional activity. It does not integrate the judicial exception into a practical application. See MPEP § 2106.05(d). Therefore, this does not amount to significantly more than the judicial exception.)
transmitting, to a remote computing device, a responsive message including at least the risk indicator and the explanatory data (This is mere data outputting, an insignificant extra solution activity, which is a well-understood, routine conventional activity. It does not integrate the judicial exception into a practical application. See MPEP § 2106.05(d). Therefore, this does not amount to significantly more than the judicial exception.)
for use in controlling access of the target entity to one or more interactive computing environments. (It only states an intended use of the transmitted information. See MPEP § 2106.05(h). Therefore, this does not amount to significantly more than the judicial exception.)
Regarding claim 2, the rejection of claim 1 is incorporated herein. Further, claim 2 recites the following abstract idea:
selecting a subset of parameter wavelet coefficients from the set of parameter wavelet coefficients that have parameter wavelet coefficients higher than remaining parameter wavelet coefficients in the set. (Comparing numerical coefficient values and selecting according to that comparison is merely mathematical calculation, which is mathematical concept. Comparison and selection process is also practical to perform in the human mind under its broadest reasonable interpretation aside from the recitation of generic computer components.)
Claim 2 does not recite additional elements.
Regarding claim 3, the rejection of claim 2 is incorporated herein. Further, claim 3 recites the following abstract idea:
wherein each parameter wavelet coefficient in the subset of parameter wavelet coefficients corresponds to a basis function in the plurality of basis functions. (Wavelet coefficients corresponding to a basis functions is merely mathematical relationship, which is mathematical concept.)
Claim 3 does not recite additional elements.
Regarding claim 4, the rejection of claim 1 is incorporated herein. Further, claim 4 recites the following additional element:
the feature learning model is a convolutional neural network configured to accept the time-series data as input and output a feature vector. (This falls under mere instructions to apply an exception. See MPEP § 2106.05(f). Therefore, this does not amount to significantly more than the judicial exception.)
Regarding claim 5, the rejection of claim 1 is incorporated herein. Further, claim 5 recites the following abstract idea:
adjusting parameters of the risk prediction model to minimize a loss function defined based on risk indicators generated for training time-series data and training risk indicators corresponding to the training time-series data. (The mathematical optimization or minimization of a loss function merely recites mathematical calculation, which is mathematical concept.)
Claim 5 does not recite additional elements.
Regarding claim 6, the rejection of claim 1 is incorporated herein. Further, claim 6 recites the following additional element:
receiving a risk assessment query for the target entity prior to applying the risk prediction model to the time-series data for the attribute associated with the target entity; (This is mere data gathering, an insignificant extra solution activity, which is a well-understood, routine conventional activity. It does not integrate the judicial exception into a practical application. See MPEP § 2106.05(d). Therefore, this does not amount to significantly more than the judicial exception.)
and accessing the attribute associated with the target entity from a database configured to store a plurality of attributes associated with a plurality of entities. (This is mere data gathering, an insignificant extra solution activity, which is a well-understood, routine conventional activity. It does not integrate the judicial exception into a practical application. See MPEP § 2106.05(d). Therefore, this does not amount to significantly more than the judicial exception.)
Regarding claim 7, the rejection of claim 1 is incorporated herein. Further, claim 7 recites the following abstract idea:
wherein the explanatory data indicates a feature of the time-series data that has a higher contribution to the risk indicator than other features of the time-series data. (Evaluating or identifying which feature contributed more is practical to perform in the human mind under its broadest reasonable interpretation aside from the recitation of generic computer components.)
Claim 7 does not recite additional elements.
Regarding claim 8, the rejection of claim 1 is incorporated herein. Further, claim 8 recites the following abstract idea:
providing a recommendation to the target entity based on the explanatory data, (Evaluating and giving recommendation is practical to perform in the human mind under its broadest reasonable interpretation aside from the recitation of generic computer components.)
wherein the recommendation indicates one or more actions for the target entity to take to improve the risk indicator. (Giving recommendations for improvement is practical to perform in the human mind under its broadest reasonable interpretation aside from the recitation of generic computer components.)
Claim 8 does not recite additional elements.
Claim 9 further recites following additional elements:
a processor; and a non-transitory computer-readable medium comprising instructions that are executable by the processor to cause the processor to perform operations comprising (This falls under mere instructions to apply an exception. See MPEP § 2106.05(f). Therefore, this does not amount to significantly more than the judicial exception.)
The rest of claim 9 recites substantially similar subject matter to claim 1 respectively and is rejected with the same rationale, mutatis mutandis.
Claims 10 – 16 recite substantially similar subject matter to claims 2-8 respectively and are rejected with the same rationale, mutatis mutandis.
Claims 17 – 20 recite substantially similar subject matter to claims 1-3, 8 respectively and are rejected with the same rationale, mutatis mutandis.
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.
Claims 1,4-9,12-17 and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Turner et al. (U.S. Pub. 2020/0134439) in view of Abuadbba et al. (NPL:”DeepiSign: Invisible Fragile Watermark to Protect the Integrity and Authenticity of CNN”), further in view of Zubarev et al. (NPL:”Adaptive neural network classifier for decoding MEG signals”).
Regarding Claim 1, Turner teaches
one or more processing devices performing operations comprising: applying a risk prediction model to time-series data for an attribute associated with a target entity to generate a risk indicator for the target entity, wherein the risk prediction model comprises ([0018] of Turner states “The risk assessment computing system 130 can include a network training server 110 for building and training a neural network 120 with the monotonic property as presented herein. The risk assessment computing system 130 can further include a risk assessment server 118 for performing risk assessment for given predictor variables 124 using the trained neural network 120.” [0034] of Turner states “FIG. 2 is a flow chart depicting an example of a process 200 for utilizing a neural network to generate risk indicators for a target entity based on predictor variables associated with the target entity. At operation 202, the process 200 involves receiving a risk assessment query for a target entity from a remote computing device, such as a computing device associated with the target entity requesting the risk assessment. The risk assessment query can also be received from a remote computing device associated with an entity authorized to request risk assessment of the target entity.” [0035] of Turner states “At operation 204, the process 200 involves accessing a neural network trained to generate risk indicator values based on input predictor variables or other data suitable for assessing risks associated with an entity… In some aspects, predictor variables can be obtained from credit files, financial records, consumer records, etc. The risk indicator can indicate a level of risk associated with the entity, such as a credit score of the entity.” )
and transmitting, to a remote computing device, a responsive message including at least the risk indicator and the explanatory data for use in controlling access of the target entity to one or more interactive computing environments. ([0038] of Turner states “At operation 208, the process 200 involves generating and transmitting a response to the risk assessment query and the response can include the risk indicator generated using the neural network. The risk indicator can be used for one or more operations that involve performing an operation with respect to the target entity based on a predicted risk associated with the target entity.” [0039] of Turner sates “For example, a customer can submit a request to access the interactive computing environment using a consumer computing system 106. Based on the request, the client computing system 104 can generate and submit a risk assessment query for the customer to the risk assessment server 118. The risk assessment query can include, for example, an identity of the customer and other information associated with the customer that can be utilized to generate predictor variables.” [0042] of Turner states ”In some implementations, the explanation codes can be generated for a subset of the predictor variables that have the highest impact on the risk indicator. For example, the risk assessment application 114 can determine a rank of each predictor variable based on an impact of the predictor variable on the risk indicator.”)
Turner does not explicitly teach that
a feature learning model configured to receive the time-series data as input and a risk classification model configured to receive output of the feature learning model and generate the risk indicator as output;
accessing parameters of the feature learning model;
applying a plurality of basis functions of a wavelet transformation on the parameters of the feature learning model to generate a set of parameter wavelet coefficients;
generating explanatory data for the risk indicator based on the set of parameter wavelet coefficients;
Zubarev teaches that
a feature learning model configured to receive the time-series data as input and a risk classification model configured to receive output of the feature learning model and generate the risk indicator as output; (Pg. 2 of Zubarev states “The proposed classifier incorporates the assumptions of the generative model described above into the discriminative neural network model (see Fig 1). The first and the second layers of the network learn spatial and temporal filters, which extract a compact representation of MEG signal features contributing to the discrimination between the classes. These features make use of spatial and temporal correlations in the data to suppress noise and to obtain sufficient separation between the simultaneously active neural sources.” Pg. 3 of Zubarev states “The mapping from the temporal convolution layer to the output is provided by a single, fully-connected layer followed by a soft-max normalization. The total number of (flat tened) inputs to this layer ninputs ¼ k*t=p, where k is the number of latent components, and t =p is a number of time points after pooling with a factor of p. This final layer outputs a vector of logits with length equal to the number of classes (m). Thus, the weight matrix of the final layer has dimensions ninputs m, with each column corresponding to the contribution of all the features to a given class.” The spatial and temporal layers correspond to the feature learning model and the fully connected softmax output layer corresponds to the classification model. In combined system with Turner, the risk indicator would have been the classification output.)
accessing parameters of the feature learning model; (Pg. 3 of Zubarev states “To identify spatial and temporal features that contribute to assignment of a given sample to a particular class, we identified the nodes of the final classification layer containing the maximum positive weights (contributions) to each particular class. Because the weights of the final layer contain information about the contribution of all the features extracted by the previous layers to each class, we can map the index of the feature that has a maximum contribution to a given class in the output layer onto the shape of the temporal convolution layer. Thus, we can identify the latent component(s) as well as the (approximate) timepoints corresponding to this most informative feature.”)
Abuadbba teaches that
applying a plurality of basis functions of a wavelet transformation on the parameters of the feature learning model to generate a set of parameter wavelet coefficients; (Pg. 2 of Abuadbba states “
PNG
media_image1.png
312
336
media_image1.png
Greyscale
we first reshape the hidden layer weights (Section 2.1) to be ready for wavelet transform as depicted in line 3 (Reshape). We then convert the weights from the spatial domain to the frequency domain using the wavelet transform (Section2.2) as shown in line 4 (Wavelet_convert).” Pg. 3 of Abuadbba states “We firstly pre-process the weights in each CNN hidden layer 𝑙𝑖 so that they can be used for wavelet transform in the next stage. To achieve this, we reshape3 them from 4D (𝑎 ×𝑏 ×𝑐 ×𝑑) to 2D (𝑟 ×𝑐) form. For example, the weights in a ResNet18 hidden 𝑙13 is reshaped from 4D (3x3x256x512) into 2D (4608 x 256)… Hiding the secret directly into hidden layer weights may yield high distortion, leading to the degrading of the model accuracy. To solve this challenge, we employ Discrete Wavelet Transform (DWT) to convert the weights from their spatial domain into the frequency domain so that the most significant coefficients are preserved to rebuild the weights after hiding… In our approach, we apply five levels of wavelet packet decomposition to each layer of a CNN model (e.g., ResNet18), which results in 32 sub-bands. A wavelet family, called Daubechies with the order 2 (𝑑𝑏2), is chosen in the transformation process because its performance in analyzing discontinuous-disturbance-dynamic signals has already been proven in [1, 20].”)
generating explanatory data for the risk indicator based on the set of parameter wavelet coefficients; (Pg. 3 of Zubarev states “To identify spatial and temporal features that contribute to assignment of a given sample to a particular class, we identified the nodes of the final classification layer containing the maximum positive weights (contributions) to each particular class. Because the weights of the final layer contain information about the contribution of all the features extracted by the previous layers to each class, we can map the index of the feature that has a maximum contribution to a given class in the output layer onto the shape of the temporal convolution layer.” Pg. 3 of Abuadbba states “We firstly pre-process the weights in each CNN hidden layer 𝑙𝑖 so that they can be used for wavelet transform in the next stage. To achieve this, we reshape3 them from 4D (𝑎 ×𝑏 ×𝑐 ×𝑑) to 2D (𝑟 ×𝑐) form. For example, the weights in a ResNet18 hidden 𝑙13 is reshaped from 4D (3x3x256x512) into 2D (4608 x 256)… Hiding the secret directly into hidden layer weights may yield high distortion, leading to the degrading of the model accuracy. To solve this challenge, we employ Discrete Wavelet Transform (DWT) to convert the weights from their spatial domain into the frequency domain so that the most significant coefficients are preserved to rebuild the weights after hiding… In our approach, we apply five levels of wavelet packet decomposition to each layer of a CNN model (e.g., ResNet18), which results in 32 sub-bands. A wavelet family, called Daubechies with the order 2 (𝑑𝑏2), is chosen in the transformation process because its performance in analyzing discontinuous-disturbance-dynamic signals has already been proven in [1, 20].” [0041] of Turner states “In other examples, the neural network can also be utilized to generate adverse action codes or other explanation codes for the predictor variables. An adverse action code can indicate an effect or an amount of impact that a given predictor variable has on the value of the credit score or other risk indicator (e.g., the relative negative impact of the predictor variable on a credit score or other risk indicator). In some aspects, the risk assessment application uses the neural network to provide adverse action codes that are compliant with regulations, business policies, or other criteria used to generate risk evaluations.” It would have been obvious to apply Abuadbba’s weight domain wavelet decomposition to the class relevant temporal filters in Zubarev and using the resulting multiscale coefficient representation from Zubarev’s parameter interpretation analysis. The resulting spectral or multiscale characterization comprises explanatory data based on parameter wavelet coefficient and turner uses the explained classification as a risk classification to connect the resulting explanation with the risk indicator.)
It would have been obvious to one with ordinary skill in the art before the effective filing date of the invention to combine the teachings of Turner, Zaburev, Abuadbba. Turner teaches using a neural network to generate a risk indicator and explanation codes for use in controlling access to an interactive computing environment. Zubarev teaches a CNN that processes time series data, supplies learned temporal features to a classifier and interprets class relevant temporal filter parameters. Abuadbba teaches applying a multilevel wavelet transformation to CNN layer weights to generate wavelet coefficients. One with ordinary skill in the art would be motivated to incorporate the teachings of Zaburev, Abuadbba into that of Turner to process time varying predictor data and explain the resulting risk classification using a multiscale representation of the learned temporal filter parameters. The combination would have been predictable to use known techniques and produce the expected risk prediction and parameter based explanation accurately.
Regarding claim 4, the rejection of claim 1 is incorporated herein. Furthermore, the combination of Turner, Abuadbba, Zubarev teach
wherein the feature learning model is a convolutional neural network configured to accept the time-series data as input and output a feature vector. (Pg. 2 of Zubarev states “The proposed classifier incorporates the assumptions of the generative model described above into the discriminative neural network model (see Fig 1). The first and the second layers of the network learn spatial and temporal filters, which extract a compact representation of MEG signal features contributing to the discrimination between the classes. These features make use of spatial and temporal correlations in the data to suppress noise and to obtain sufficient separation between the simultaneously active neural sources.” Pg. 3 of Zubarev states “We used two variants of this layer. The simpler one (LF-CNN) applies k separate 1-dimensional convolution filters of the l-th order to the time courses of the k spatial components produced by the input layer. The model assumes that these time courses do not interact and that they have unique spectral fingerprints. This layer variant can be viewed as applying linear finite-impulse-response filters (hence LF) that specifically capture the informative features in time courses of each spatial component…The mapping from the temporal convolution layer to the output is provided by a single, fully-connected layer followed by a soft-max normalization. The total number of (flattened) inputs to this layer ninputs ¼ k*t=p, where k is the number of latent components, and t =p is a number of time points after pooling with a factor of p.” The flattened compact feature representation is a feature vector supplied by the feature learning portion to the classifier.)
Regarding claim 5, the rejection of claim 1 is incorporated herein. Furthermore, the combination of Turner, Abuadbba, Zubarev teach
adjusting parameters of the risk prediction model to minimize a loss function defined based on risk indicators generated for training time-series data and training risk indicators corresponding to the training time-series data. ([0044] of Turner states “As illustrated in FIG. 4, the training samples 402 can include multiple training vectors consisting of training predictor variables and training outputs, i.e. training risk indicators. A particular training vector i can include an N-dimensional input predictor vector X(i)=[xi (i), . . . , xN-1 (i), 1] constituting particular values of the training predictor variables, where i=1, . . . , T and T is the number of training vectors in the training samples. The particular training vector i can also include a training output z(i), i.e. a training risk indicator or outcome corresponding to the input predictor vector X(i).” [0050] of Turner states “Referring back to FIG. 3, the process 300 involves formulating an optimization problem for the neural network model at operation 306. Training a neural network can include solving an optimization problem to find the parameters of the neural network, such as the weights of the connections in the neural network. In particular, training the neural network 400 can involve determining the values of the weights w in the neural network 400, i.e. w(0), w(1), and w(2), so that a loss function L(w) of the neural network 400 is minimized. The loss function can be defined as, or as a function of, the difference between the outputs predicted using the neural network with weights w, denoted as {circumflex over (Z)}=[{circumflex over (z)}(1) {circumflex over (z)}(2) . . . {circumflex over (z)}(T)] and the observed output Z=[z(1) z(2) . . . z(T)]. ” Pg. 5 of Zubarev states “We initialized the bias variables to a constant value of 0.1. We used the Adam optimization algorithm with a batch size of 100 and the learning rate of 3:0 _ 10_4 to optimize multinomial cross-entropy between the model predictions and true labels. Higher learning rates were also used but they did not improve performance. We used an early-stopping strategy to prevent over-fitting; for every 1000 iterations, we computed the validation cost (multinomial cross-entropy) and stopped the iterations immediately if the cost function value was increasing or decreasing by less than 1:0 _ 10_5.” )
Regarding claim 6, the rejection of claim 1 is incorporated herein. Furthermore, the combination of Turner, Abuadbba, Zubarev teach
receiving a risk assessment query for the target entity prior to applying the risk prediction model to the time-series data for the attribute associated with the target entity; ([0034] of Turner states “FIG. 2 is a flow chart depicting an example of a process 200 for utilizing a neural network to generate risk indicators for a target entity based on predictor variables associated with the target entity. At operation 202, the process 200 involves receiving a risk assessment query for a target entity from a remote computing device, such as a computing device associated with the target entity requesting the risk assessment. The risk assessment query can also be received from a remote computing device associated with an entity authorized to request risk assessment of the target entity.”)
and accessing the attribute associated with the target entity from a database configured to store a plurality of attributes associated with a plurality of entities. ([0037] of Turner states “At operation 206, the process 200 involves applying the neural network to generate a risk indicator for the target entity specified in the risk assessment query. Predictor variables associated with the target entity can be used as inputs to the neural network. The predictor variables associated with the target entity can be obtained from a predictor variable database configured to store predictor variables associated with various entities. The output of the neural network would include the risk indicator for the target entity based on its current predictor variables.”)
Regarding claim 7, the rejection of claim 1 is incorporated herein. Furthermore, the combination of Turner, Abuadbba, Zubarev teach
wherein the explanatory data indicates a feature of the time-series data that has a higher contribution to the risk indicator than other features of the time-series data. ([0042] of Turner states “In some implementations, the explanation codes can be generated for a subset of the predictor variables that have the highest impact on the risk indicator. For example, the risk assessment application 114 can determine a rank of each predictor variable based on an impact of the predictor variable on the risk indicator. A subset of the predictor variables including a certain number of highest-ranked predictor variables can be selected and explanation codes can be generated for the selected predictor variables. The risk assessment application 114 may provide recommendations to a target entity based on the generated explanation codes. The recommendations may indicate one or more actions that the target entity can take to improve the risk indicator (e.g., improve a credit score).” Pg. 3 of Zubarev states “To identify spatial and temporal features that contribute to assignment of a given sample to a particular class, we identified the nodes of the final classification layer containing the maximum positive weights (contributions) to each particular class. Because the weights of the final layer contain information about the contribution of all the features extracted by the previous layers to each class, we can map the index of the feature that has a maximum contribution to a given class in the output layer onto the shape of the temporal convolution layer.” In combination, Zubarev and Turner teaches explanatory data identifying a time series features having a higher contribution to the risk indicator than other time series features.)
Regarding claim 8, the rejection of claim 1 is incorporated herein. Furthermore, the combination of Turner, Abuadbba, Zubarev teach
providing a recommendation to the target entity based on the explanatory data, wherein the recommendation indicates one or more actions for the target entity to take to improve the risk indicator. ([0042] of Turner states “In some implementations, the explanation codes can be generated for a subset of the predictor variables that have the highest impact on the risk indicator. For example, the risk assessment application 114 can determine a rank of each predictor variable based on an impact of the predictor variable on the risk indicator. A subset of the predictor variables including a certain number of highest-ranked predictor variables can be selected and explanation codes can be generated for the selected predictor variables. The risk assessment application 114 may provide recommendations to a target entity based on the generated explanation codes. The recommendations may indicate one or more actions that the target entity can take to improve the risk indicator (e.g., improve a credit score).”)
Claims 9,12-16 recite substantially similar subject matter to claims 1, 4-8 respectively and are rejected with the same rationale, mutatis mutandis.
Claims 17, 20 recite substantially similar subject matter to claims 1, 8 respectively and are rejected with the same rationale, mutatis mutandis.
Claims 2, 10, 18 are rejected under 35 U.S.C. 103 as being unpatentable over Turner et al. (U.S. Pub. 2020/0134439) in view of Abuadbba et al. (NPL:”DeepiSign: Invisible Fragile Watermark to Protect the Integrity and Authenticity of CNN”), Zubarev et al. (NPL:”Adaptive neural network classifier for decoding MEG signals”), further in view of Donoho et al. (U.S. Pub. 6766062).
Regarding claim 2, the rejection of claim 1 is incorporated herein. Furthermore, the combination of Turner, Abuadbba, Zubarev does not teach
wherein the operations further comprise selecting a subset of parameter wavelet coefficients from the set of parameter wavelet coefficients that have parameter wavelet coefficients higher than remaining parameter wavelet coefficients in the set.
Donoho teaches that
wherein the operations further comprise selecting a subset of parameter wavelet coefficients from the set of parameter wavelet coefficients that have parameter wavelet coefficients higher than remaining parameter wavelet coefficients in the set. (Column 1 Lines 62 – Column 2 Lines 4 of Donoho states that “A Wavelet transform is performed on values derived from the frequency domain values provided in digital polar coordinates to generate Wavelet coefficients (or Ridgelet coefficients). Next a thresholding process can be performed. According to the thresholding process, the Wavelet coefficients are filtered to select a group of larger Wavelet coefficients and discard the remaining Wavelet coefficients (e.g., Select those coefficients which are greater than a threshold, and discard the remaining coefficients).)
It would have been obvious to one with ordinary skill in the art before the effective filing date of the invention to combine the teachings of Donoho with the combination of Turner, Zaburev, Abuadbba. Turner teaches using a neural network to generate a risk indicator and explanation codes for use in controlling access to an interactive computing environment. Zubarev teaches a CNN that processes time series data, supplies learned temporal features to a classifier and interprets class relevant temporal filter parameters. Abuadbba teaches applying a multilevel wavelet transformation to CNN layer weights to generate wavelet coefficients. Donoho teaches selecting larger wavelet coefficients and discarding the remaining smaller coefficients. One with ordinary skill in the art would be motivated to incorporate the teachings of Donoho into that of Turner, Zubarev, Abuadbba so that the explanation focuses on the coefficients representing the dominant components of the learned temporal filter parameters while excluding smaller coefficients. The combination would have been predictable to apply known selection technique to the wavelet coefficients generated from the CNN weights for improved accuracy.
Claims 10, 18 recite substantially similar subject matter to claim 2 respectively and are rejected with the same rationale, mutatis mutandis.
Claims 3, 11, 19 are rejected under 35 U.S.C. 103 as being unpatentable over Turner et al. (U.S. Pub. 2020/0134439) in view of Abuadbba et al. (NPL:”DeepiSign: Invisible Fragile Watermark to Protect the Integrity and Authenticity of CNN”), Zubarev et al. (NPL:”Adaptive neural network classifier for decoding MEG signals”), Donoho et al. (U.S. Pub. 6766062), further in view of Wolter et al. (NPL:”Neural Network Compression via Learnable Wavelet Transforms”)
Regarding claim 3, the rejection of claim 2 is incorporated herein. Furthermore, the combination of Turner, Abuadbba, Zubarev, Donoho does not teach
wherein each parameter wavelet coefficient in the subset of parameter wavelet coefficients corresponds to a basis function in the plurality of basis functions.
Wolter teaches that
wherein each parameter wavelet coefficient in the subset of parameter wavelet coefficients corresponds to a basis function in the plurality of basis functions. (Pg. 2 of Wolter states “We advocate the use of wavelets as an alternative for representing the weight matrices of linear layers. Using wavelets offers us two key advantages. Firstly, we can apply the fast wavelet transform (FWT), which has only a complexity of O(n) for projection and comes with a large selection of possible basis functions. Secondly, we can build upon the product filter approach for wavelet design [12] to directly integrate the learning of wavelet bases as a part of training CNNs or RNNs. Learning the bases gives us added flexibility in representing their weight matrices. Motivated by these advantages, we propose a new linear layer which directly integrates the FWT into its formulation so layer weights can be represented as sparse wavelet coefficients. Furthermore, rather than limit ourselves to predefined wavelets as basis functions, we learn the bases directly as a part of network training.” Pg. 3 of Wolter states “For our purposes, we can consider the wavelet transform as being analogous to the Fourier transform. Similarly, wavelets are akin to sinusoids, with a key distinction however, that wavelets are localized basis functions, i.e. are not infinite… Given a signal x indexed by n, the forward wavelet transform yields coefficients bjk in vector b, which projects x onto the wavelet basis A.”)
It would have been obvious to one with ordinary skill in the art before the effective filing date of the invention to combine the teachings of Wolter with the combination of Turner, Zaburev, Abuadbba, Donoho. Turner teaches using a neural network to generate a risk indicator and explanation codes for use in controlling access to an interactive computing environment. Zubarev teaches a CNN that processes time series data, supplies learned temporal features to a classifier and interprets class relevant temporal filter parameters. Abuadbba teaches applying a multilevel wavelet transformation to CNN layer weights to generate wavelet coefficients. Donoho teaches selecting larger wavelet coefficients and discarding the remaining smaller coefficients. Wolter teaches representing neural network weights using wavelet bases and corresponding coefficients where the coefficients also identify respective scale and time positions. One with ordinary skill in the art would be motivated to incorporate the teachings of Wolter into that of Turner, Zubarev, Abuadbba, Donoho so that each selected parameter wavelet coefficient identifies the corresponding wavelet basis function and its represented scale and position. The combination would have been predictable use of the known basis function correspondence for the wavelet coefficients selected.
Claims 11, 19 recite substantially similar subject matter to claim 3 respectively and are rejected with the same rationale, mutatis mutandis.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to BYUNGKWON HAN whose telephone number is (571)272-5294. The examiner can normally be reached M-F: 9:00AM-6PM PST.
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, Li B Zhen can be reached at (571)272-3768. 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.
/BYUNGKWON HAN/ Examiner, Art Unit 2121
/Li B. Zhen/ Supervisory Patent Examiner, Art Unit 2121
Prosecution Insights