Detailed Action
This action is in response to the application filed 10/10/2022, in which:
Claims 1, 9 and 17 are the independent claims.
Claims 1-20 are currently pending.
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 Objections
Claims 1-20 are objected to because of the following informalities: ‘neutral’ is recited within Claims 1-2, 9-10, and 17-18. ‘leafs’ is recited within Claims 1, 9, and 17. ‘descried’ is recited within Claims 10 and 11. All three terms appear to be typographical errors. Appropriate correction is required. For the purposes of examination, the examiner interprets ‘neutral’ as ‘neural’ and ‘leafs’ as ‘leaves’. ‘descried’ as ‘described’.
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.
Regarding Claim 1:
Subject Matter Eligibility Analysis Step 1:
Claim 1 recites a method, thus a process, one of the four statutory categories of patentable subject matter.
Subject Matter Eligibility Analysis Step 2A Prong 1:
However, Claim 1 further recites the method comprising:
initializing a decision tree, and setting a root of the decision tree (a human being can mentally apply evaluation to initialize a decision tree and set a root for the decision tree)
branching leafs from the root of the decision tree based on effective filters of the neutral network as a decision rule, until all effective filters of the neutral network are covered by the decision tree (a human being, with the aid of pen and paper, can mentally apply evaluation to create a decision tree by connecting leaves from the root of the tree to leaves until all filters of a model are covered by the created decision tree)
Claim 1 thus recites an abstract idea (that falls into the “mental processes” group of abstract ideas).
Subject Matter Eligibility Analysis Step 2A Prong 2:
This judicial exception is not integrated into a practical application because the additional elements recited consists of:
A method for converting neural network, applied to a terminal device, comprising: (which is restricting the abstract idea to a Particular Technological Environment, by MPEP 2106.05(h))
wherein the neutral network is a piece-wise linearly activated neutral network (which is restricting the abstract idea to a Particular Technological Environment, by MPEP 2106.05(h))
Subject Matter Eligibility Analysis Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because the additional elements recited, alone or in combination, do not provide significantly more than the abstract idea itself. Additional elements a and b are only restricting the abstract idea to a Particular Technological Environment (MPEP 2106.05(h)) which cannot provide significantly more. Thus, the claim is subject-matter ineligible.
Regarding Claim 2:
Subject Matter Eligibility Analysis Step 1:
Dependent Claim 2 recites the method of Claim 1. Claim 1 is a method, thus a process, one of the four statutory categories of patentable subject matter.
Subject Matter Eligibility Analysis Step 2A Prong 1:
However, Claim 2 does not recite any additional abstract ideas and only inherits the abstract ideas from Claim 1. Claim 2 thus recites an abstract idea.
Subject Matter Eligibility Analysis Step 2A Prong 2:
This judicial exception is not integrated into a practical application because the additional elements recited consists of:
wherein the branching leafs from the root of the decision tree comprises: starting from nodes branched from the root of the decision tree, further branching the nodes into leaf branches each corresponding to an effective filter (which is restricting the abstract idea to a Particular Technological Environment, by MPEP 2106.05(h))
wherein an order of the effective filters is based on an order of the effective filters in a same layer of the neutral network and orders in different layers of the neutral network (which is restricting the abstract idea to a Particular Technological Environment, by MPEP 2106.05(h))
Subject Matter Eligibility Analysis Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because the additional elements recited, alone or in combination, do not provide significantly more than the abstract idea itself. Additional elements a and b are only restricting the abstract idea to a Particular Technological Environment (MPEP 2106.05(h)) which cannot provide significantly more. Thus, the claim is subject-matter ineligible.
Regarding Claim 3:
Subject Matter Eligibility Analysis Step 1:
Dependent Claim 3 recites the method of Claim 1. Claim 3 is a method, thus a process, one of the four statutory categories of patentable subject matter.
Subject Matter Eligibility Analysis Step 2A Prong 1:
However, Claim 3 further recites the method comprising wherein, for a fully connected layer, an effective matrix is adopted as the decision rule (a mathematical relationship between variables and/or numbers using a mathematical formula/equations). Claim 3 thus recites an abstract idea (that falls into the “mental processes” group of abstract ideas).
Subject Matter Eligibility Analysis Step 2A Prong 2:
This judicial exception is not integrated into a practical application because there are no new additional elements recited.
Subject Matter Eligibility Analysis Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because there are no new additional elements recited. The judicial exception alone does not provide significantly more than the abstract idea itself. Thus, the claim is subject-matter ineligible.
Regarding Claim 4:
Subject Matter Eligibility Analysis Step 1:
Dependent Claim 4 recites the method of Claim 1. Claim 4 is a method, thus a process, one of the four statutory categories of patentable subject matter.
Subject Matter Eligibility Analysis Step 2A Prong 1:
However, Claim 4 further recites the method comprising wherein, for a skip connection layer, a residual effective matrix is adopted as the decision rule (a mathematical relationship between variables and/or numbers using a mathematical formula/equations). Claim 4 thus recites an abstract idea (that falls into the “mathematical concepts” group of abstract ideas).
Subject Matter Eligibility Analysis Step 2A Prong 2:
This judicial exception is not integrated into a practical application because there are no new additional elements recited.
Subject Matter Eligibility Analysis Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because there are no new additional elements recited. The judicial exception alone does not provide significantly more than the abstract idea itself. Thus, the claim is subject-matter ineligible.
Regarding Claim 5:
Subject Matter Eligibility Analysis Step 1:
Dependent Claim 5 recites the method of Claim 1. Claim 5 is a method, thus a process, one of the four statutory categories of patentable subject matter.
Subject Matter Eligibility Analysis Step 2A Prong 1:
However, Claim 5 further recites the method comprising wherein, for a normalization layer, the normalization layer is embedded in a linear layer before or after pre-activation normalization or post-activation normalization, respectively (a mathematical relationship between variables and/or numbers using a mathematical formula/equations). Claim 5 thus recites an abstract idea (that falls into the “mathematical concepts” group of abstract ideas).
Subject Matter Eligibility Analysis Step 2A Prong 2:
This judicial exception is not integrated into a practical application because there are no new additional elements recited.
Subject Matter Eligibility Analysis Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because there are no new additional elements recited. The judicial exception alone does not provide significantly more than the abstract idea itself. Thus, the claim is subject-matter ineligible.
Regarding Claim 6:
Subject Matter Eligibility Analysis Step 1:
Dependent Claim 6 recites the method of Claim 1. Claim 6 is a method, thus a process, one of the four statutory categories of patentable subject matter.
Subject Matter Eligibility Analysis Step 2A Prong 1:
However, Claim 6 further recites the method comprising wherein, for a convolution layer, an effective convolution is adopted as the decision rule (a mathematical relationship between variables and/or numbers using a mathematical formula/equations). Claim 6 thus recites an abstract idea (that falls into the “mathematical concepts” group of abstract ideas).
Subject Matter Eligibility Analysis Step 2A Prong 2:
This judicial exception is not integrated into a practical application because there are no new additional elements recited.
Subject Matter Eligibility Analysis Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because there are no new additional elements recited. The judicial exception alone does not provide significantly more than the abstract idea itself. Thus, the claim is subject-matter ineligible.
Regarding Claim 7:
Subject Matter Eligibility Analysis Step 1:
Dependent Claim 7 recites the method of Claim 1. Claim 7 is a method, thus a process, one of the four statutory categories of patentable subject matter.
Subject Matter Eligibility Analysis Step 2A Prong 1:
However, Claim 7 further recites the method comprising lossless pruning the decision tree based on violating rules and/or redundant rules of the decision tree (a mathematical relationship between variables and/or numbers using a mathematical formula/equations). Claim 7 thus recites an abstract idea (that falls into the “mathematical concepts” group of abstract ideas).
Subject Matter Eligibility Analysis Step 2A Prong 2:
This judicial exception is not integrated into a practical application because there are no new additional elements recited.
Subject Matter Eligibility Analysis Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because there are no new additional elements recited. The judicial exception alone does not provide significantly more than the abstract idea itself. Thus, the claim is subject-matter ineligible.
Regarding Claim 8:
Subject Matter Eligibility Analysis Step 1:
Dependent Claim 8 recites the method of Claim 1. Claim 8 is a method, thus a process, one of the four statutory categories of patentable subject matter.
Subject Matter Eligibility Analysis Step 2A Prong 1:
However, Claim 8 further recites the method comprising lossless pruning the decision tree based on categories realized during training of the neural network (a mathematical relationship between variables and/or numbers using a mathematical formula/equations). Claim 8 thus recites an abstract idea (that falls into the “mathematical concepts” group of abstract ideas).
Subject Matter Eligibility Analysis Step 2A Prong 2:
This judicial exception is not integrated into a practical application because there are no new additional elements recited.
Subject Matter Eligibility Analysis Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because there are no new additional elements recited. The judicial exception alone does not provide significantly more than the abstract idea itself. Thus, the claim is subject-matter ineligible.
Regarding Claims 9-16:
Claims 9-16 incorporate substantively all the limitations of Claims 1-8 in an electronic device (thus, a machine) and further recites a memory storing executable instructions; and at least one processor coupled to the memory, wherein when executing the executable instructions, the at least one processor is configured to: (these claim limitations appear to perform a mental process and the performance of an abstract idea on a computer is no more than instructions to “apply it” on a computer, by MPEP 2106.05(f)) and does not appear to integrate the abstract idea into a particular application; thus, the claim is subject-matter ineligible as it does not include additional elements that are sufficient to amount to significantly more than the judicial exception because the additional elements, alone or in combination, do not provide significantly more than the abstract idea itself); thus, Claims 9-16 are rejected for reasons set forth in the rejections of Claims 1-8, respectively.
Regarding Claims 17-20:
Claims 17-20 incorporate substantively all the limitations of Claims 1-2 and 7-8 in an non-transitory storage medium storing computer executable instructions (thus, a manufacture) and further recites wherein when the computer executable instructions are executed on a computer, the computer is triggered to: (these claim limitations appear to perform a mental process and the performance of an abstract idea on a computer is no more than instructions to “apply it” on a computer, by MPEP 2106.05(f)) and does not appear to integrate the abstract idea into a particular application; thus, the claim is subject-matter ineligible as it does not include additional elements that are sufficient to amount to significantly more than the judicial exception because the additional elements, alone or in combination, do not provide significantly more than the abstract idea itself); thus, Claims 17-20 are rejected for reasons set forth in the rejections of Claims 1-2 and 7-8, respectively.
Claim Rejections - 35 USC § 102
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
Claims 1-3, 6, 9-11, 14, and 17-18 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Zhang et al., “Interpreting CNNs via Decision Trees”.
Regarding Claim 1:
Zhang teaches:
A method for converting neural network, applied to a terminal device, comprising:
(Zhang, Page 6261, Figure 2; Page 6265, Algorithm 1, Column 2, Paragraph 4, “…Algorithm 1 shows the pseudo-code of the learning process … This objective penalizes the decrease of the discriminative power and forces the system to summarize a few generic decision modes for explanation …”. Figure 2 denotes the decision tree creation process to emulate the CNN for prediction explainability; thus, a method for converting a neural network (CNN) into a decision tree which is applied to a terminal device (where the terminal device is interpreted by the examiner as a system/device capable of implementing the code/methods/algorithms of learning decision trees)).
initializing a decision tree, and setting a root of the decision tree; and
(Zhang, Page 6265, Algorithm 1: “Initialize a tree Q = P0 and set t = 0”; Figure 4. Algorithm 1 shows the initialization of the decision tree and setting a root of the decision tree (P0 which can also be seen within Figure 4)).
branching leafs from the root of the decision tree based on effective filters of the neutral network as a decision rule, until all effective filters of the neutral network are covered by the decision tree, wherein the neutral network is a piece-wise linearly activated neutral network.
(Zhang, Page 6265, Algorithm 1: “
PNG
media_image1.png
56
352
media_image1.png
Greyscale
”; Figure 4, Column 1, Paragraph 3, “… using which filters/parts for prediction.
PNG
media_image2.png
29
77
media_image2.png
Greyscale
(3) Different dimensions of the vector x measure the scalar signal strength of different object parts, since a filter potentially represents a certain object part. g corresponds to the selection of object parts for the CNN prediction”; Page 6264, Column 2, Paragraph 5, “… we can use a piecewise linear representation to represent the function of cascaded FC layers and ReLU layers, as follows …”; Page 6262, Figure 2; Page 6263, Paragraph 2, “Each node (decision mode) in the parse tree quantitatively explains the prediction at a different abstraction level, i.e. clarifying how much each object part/filter contributes to the prediction score … decision modes usually select significant object parts (filters) …”. Figure 4 shows the child of the root nodes (interpreted by the examiner as leaves) which are connected from the root to the child nodes/leaf via a branch. These child nodes/leaves denoted as gi within Zhang which corresponds to a partial part of the object/image; thus, interpreted as an effective filter by the examiner as each gi node represents a decision mode (which is interpreted by the examiner as a type of decision rule as they represent selecting the significant filters (effective filters)). The nodes are added until all effective filters of the neural network are covered by the decision tree (Algorithm 1:
PNG
media_image3.png
22
38
media_image3.png
Greyscale
)).
Regarding Claim 2:
Zhang teaches the method of Claim 1 and further teaches:
wherein the branching leafs from the root of the decision tree comprises:
starting from nodes branched from the root of the decision tree, further branching the nodes into leaf branches each corresponding to an effective filter, wherein an order of the effective filters is based on an order of the effective filters in a same layer of the neutral network and orders in different layers of the neutral network.
(Zhang, Page 6262, Figure 2; Page 6265, Figure 4. Figure 4 shows the further connections between nodes corresponding to the Filters (1->D); where each effective filter represents a certain object part (and is better depicted within Figure 2). The effective filters within Zhang are layer-specific as each layer is associated to a different object/part and items within the same layer are under the same categorical object/part. Figure 4 further shows the effective filters corresponding contributing scores being mapped (interpreted by the examiner as being based on an order) to the w matrix that correspond to the different layers and a same layer within the decision trees (which is depicted within Figure 4’s decisions trees indicating the branching/learning process)).
Regarding Claim 3:
Zhang teaches the method of Claim 1 and further teaches:
wherein, for a fully connected layer, an effective matrix is adopted as the decision rule.
(Zhang, Page 6262, Figure 2: ‘FC Layers’; Page 6264, Column 2, Paragraph 3-5, “… we choose filters in the top conv-layer to represent object parts. Consequently, we quantitatively analyze how fully-connected (FC) layers use object-part features from the top conv-layer to make final predictions … we can use a piecewise linear representation to represent the function of cascaded FC layers and ReLU layers …”; Page 6266, Column 1, Paragraph 6, “… We compute the vector … to evaluate the contribution of different filters and that of different object parts.
PNG
media_image4.png
20
189
media_image4.png
Greyscale
(11) … we label a matrix A ∈ {0, 1}M×D to assign each filter in the top conv-layer with a specific object part … ”; Page 6265, Algorithm 1: Assign filters with semantic object parts to obtain A. The fully connected layers (FC) are used to make the final predictions based on the decision modes (rules) as the object-part features from the top conv-layer to make final predictions. Equation 11 shows the use of an effective matrix A within
PNG
media_image5.png
21
65
media_image5.png
Greyscale
which indicates whether a filter should be assigned (0 or 1) to a specific part; thus, a final decision rule which uses the effective matrix created from Algorithm 1).
Regarding Claim 6:
Zhang teaches the method of Claim 1 and further teaches:
wherein, for a convolution layer, an effective convolution is adopted as the decision rule.
(Zhang, Page 6262, Figure 2: ‘top conv-layer’; Page 6265, Figure 4. Figure 2 and 4 both indicate the conv-layer (convolutional layer) filter being adopted as decision rules (modes) within the tree which is based in the effective filter(object part)).
Regarding Claims 9-11 and 14:
Claims 9-11 and 14 incorporate substantively all the limitations of Claims 1-3 and 6 in an electronic device and further recites a memory storing executable instructions; and at least one processor coupled to the memory, wherein when executing the executable instructions, the at least one processor is configured to: (Zhang, Page 6261, Figure 2; Page 6265, Algorithm 1, Column 2, Paragraph 4, “…Algorithm 1 shows the pseudo-code of the learning process … This objective penalizes the decrease of the discriminative power and forces the system to summarize a few generic decision modes for explanation …”. The examiner interprets the system/electronic device capable of implementing the code/methods/algorithms of learning decision trees; thus, the device contains a memory storing executable instructions; and at least one processor coupled to the memory, wherein when executing the executable instructions, the at least one processor is configured to implement the code/methods/algorithms of learning decision trees)); thus, Claims 9-11 and 14 are rejected for reasons set forth in the rejections of Claims 1-3 and 6, respectively.
Regarding Claims 17 and 18:
Claims 17 and 18 incorporate substantively all the limitations of Claims 1 and 2 in an non-transitory storage medium storing computer executable instructions and further recites wherein when the computer executable instructions are executed on a computer, the computer is triggered to: (Zhang, Page 6261, Figure 2; Page 6265, Algorithm 1, Column 2, Paragraph 4, “…Algorithm 1 shows the pseudo-code of the learning process … This objective penalizes the decrease of the discriminative power and forces the system to summarize a few generic decision modes for explanation …”. The examiner interprets the system/electronic device capable of implementing the code/methods/algorithms of learning decision trees; thus, the device contains The decision tree learning pipeline, trains, optimizes and evaluates on a computing device in which the CPU (processor), memory, and non-transitory storage medium are inherent)); thus, Claims 17 and 18 are rejected for reasons set forth in the rejections of Claims 1 and 2, respectively.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claims 4 and 12 are rejected under 35 U.S.C. 103 as being unpatentable over Zhang et al., “Interpreting CNNs via Decision Trees”, in view of Wang et al., “Nonparametric Learning of Two-Layer ReLU Residual Units”.
Regarding Claim 4:
Zhang teaches the method of Claim 1. Zhang denotes the filter loss in the explainer is not compatible with skip connections within the residual network. Thus, Zhang fails to explicitly disclose:
wherein, for a skip connection layer, a residual effective matrix is adopted as the decision rule.
However, Wang teaches:
wherein, for a skip connection layer, a residual effective matrix is adopted as the decision rule.
(Wang, Page 2, Paragraph 1-2, “Residual networks, or ResNets … are a class of deep neural networks that adopt skip connections … In this paper, we propose algorithms that can learn a general class of single-skip two-layer residual units with ReLU activation as shown on the right by equation: y = B [(Ax)+ + x ] (Equation 1)…”. Wang teaches taking a skip connection layer such as a residual network and considering it in the form of Equation 1. Equation 4 is the decision rule for dictating the ground-truth model where the residual effective matrix is [(Ax)+ + x ] which comprises the residual weight matrix and skip-identity matrix).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to utilize Zhang’s method of converting piece-wise linearly activated neural networks into decision trees with the way Wang handles skip connect layers with a residual effect matrix for decisions as this lead to robustness due to being able to handle more architectures, efficiency due to skipping, handling different input distributions, and expressing ground-truths (Wang, Page 10, Paragraph 2-3, “…we address the problem of learning a general class of two-layer residual units and propose an algorithm based on landscape design and convex optimization: First, minimizers of our objective functionals can express the exact ground-truth network … Moreover, our algorithms solving both layers and the whole networks are strongly consistent, with very weak conditions on input distributions. Our work opens the door to a variety of open problems to explore. We provide a strong consistency result without assuming a particular input distribution …”).
Regarding Claim 12:
Claim 12 incorporates substantively all the limitations of Claim 4 in an electronic device and further recites a memory storing executable instructions; and at least one processor coupled to the memory, wherein when executing the executable instructions, the at least one processor is configured to: (Zhang, Page 6261, Figure 2; Page 6265, Algorithm 1, Column 2, Paragraph 4, “…Algorithm 1 shows the pseudo-code of the learning process … This objective penalizes the decrease of the discriminative power and forces the system to summarize a few generic decision modes for explanation …”. The examiner interprets the system/electronic device capable of implementing the code/methods/algorithms of learning decision trees; thus, the device contains a memory storing executable instructions; and at least one processor coupled to the memory, wherein when executing the executable instructions, the at least one processor is configured to implement the code/methods/algorithms of learning decision trees)); thus, Claim 12 is rejected for reasons set forth in the rejections of Claim 4.
Claims 5 and 13 are rejected under 35 U.S.C. 103 as being unpatentable over Zhang et al., “Interpreting CNNs via Decision Trees”, in view of Darma et al., “The Regularization Effect of Pre-activation Batch Normalization on Convolutional Neural Network Performance for Face Recognition System Paper”.
Regarding Claim 5:
Zhang teaches the method of Claim 1. Zhang fails to explicitly disclose:
wherein, for a normalization layer, the normalization layer is embedded in a linear layer before … pre-activation normalization …
However, Darma teaches:
wherein, for a normalization layer, the normalization layer is embedded in a linear layer before … pre-activation normalization …
(Darma, Page 301, Figure 1. Fig. 1 shows the structure of the pre-activation batch normalization for the CNN; thus, the normalization layer is embedded in a linear layer before pre-activation normalization).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to utilize Zhang’s method of converting piece-wise linearly activated neural networks into decision trees with the way Darma handles normalization layers due to being able to improve classification, increase performance within facial recognition, and stabilization (Darma, Page 309, Paragraph 1, “…In this research, the best deep learning CNN architecture for recognizing human face images was obtained from the Pre-Activation-BN-CNN Architecture. This model is powered by a batch normalization layer at each of its four convolutional units. The batch normalization layers are all placed before the rectified linear units (Relu) activation function. The result of this research shows that placing the Batch Normalization layer before the ReLu activation function improved network classification power, as the model training and validation results showed 100.00% and 99.87%, respectively. This shows the regularization effect of the Pre-Activation-BN-CNN model over the other three CNN architectures for face recognition systems. In this research work, the application of the Pre- Activation-BN-CNN Architecture has enhanced the performance of the face recognition system … This leads to better network convergence. The experiment results also show that having the rectified linear unit (Relu) activation function in a model architecture stabilizes network training and improves model classification …”).
Regarding Claim 13:
Claim 13 incorporates substantively all the limitations of Claim 5 in an electronic device and further recites a memory storing executable instructions; and at least one processor coupled to the memory, wherein when executing the executable instructions, the at least one processor is configured to: (Zhang, Page 6261, Figure 2; Page 6265, Algorithm 1, Column 2, Paragraph 4, “…Algorithm 1 shows the pseudo-code of the learning process … This objective penalizes the decrease of the discriminative power and forces the system to summarize a few generic decision modes for explanation …”. The examiner interprets the system/electronic device capable of implementing the code/methods/algorithms of learning decision trees; thus, the device contains a memory storing executable instructions; and at least one processor coupled to the memory, wherein when executing the executable instructions, the at least one processor is configured to implement the code/methods/algorithms of learning decision trees)); thus, Claim 13 is rejected for reasons set forth in the rejections of Claim 5.
Claims 7-8, 15-16, and 19-20 are rejected under 35 U.S.C. 103 as being unpatentable over Zhang et al., “Interpreting CNNs via Decision Trees”, in view of Izza et al., “On Tackling Explanation Redundancy in Decision Trees”.
Regarding Claim 7:
Zhang teaches the method of Claim 1. Zhang fails to explicitly disclose:
lossless pruning the decision tree based on violating rules and/or redundant rules of the decision tree.
However, Izza teaches:
lossless pruning the decision tree based on … redundant rules of the decision tree.
(Izza, Page 264, Main Results #7: “The paper offers extensive experimental evidence, attesting to the significance of identifying and removing explanation redundancy from decision tree paths”. Izza teaches lossless pruning via removing redundant decision tree paths (interpreted as rules) which is a form of lossless optimization as information is not lost and the simplified tree produces the same output).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to utilize Zhang’s method of converting piece-wise linearly activated neural networks into decision trees with the way Izza handles pruning decision trees based on redundancy to optimize decision making, improve performance + efficiency, and analysis of irrelevant features (Izza, Page 309, Paragraph 1, “…This paper investigates path explanation redundancy in decision trees, i.e. the existence of features that are irrelevant for the prediction associated with a given path. In addition, the paper also shows that the computation of irredundant path explanations in DTs is tightly related with recent work on computing abductive explanations …”).
Regarding Claim 8:
Zhang teaches the method of Claim 1. Zhang fails to explicitly disclose:
lossless pruning the decision tree based on categories realized during training of the neural network.
However, Izza teaches:
lossless pruning the decision tree based on categories realized during training of the neural network.
(Izza, Page 263-264, Main Results #1: “The paper formalizes in detail the computation of explanations in decision trees, such that decision trees are allowed to have both categorical and ordinal features …” & #7: “The paper offers extensive experimental evidence, attesting to the significance of identifying and removing explanation redundancy from decision tree paths”; Page 273, Figure 3. Figure 3 shows the categorical decision tree that is based on categorical features and realized and learned through the training process; where the removal of redundant paths takes into consideration categories/features realized during the training).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to utilize Zhang’s method of converting piece-wise linearly activated neural networks into decision trees with the way Izza handles pruning decision trees based on redundancy to optimize decision making, improve performance + efficiency, and analysis of irrelevant features (Izza, Page 309, Paragraph 1, “…This paper investigates path explanation redundancy in decision trees, i.e. the existence of features that are irrelevant for the prediction associated with a given path. In addition, the paper also shows that the computation of irredundant path explanations in DTs is tightly related with recent work on computing abductive explanations …”).
Regarding Claim 15 and 16:
Claims 15 and 16 incorporate substantively all the limitations of Claims 7 and 8 in an electronic device and further recites a memory storing executable instructions; and at least one processor coupled to the memory, wherein when executing the executable instructions, the at least one processor is configured to: (Zhang, Page 6261, Figure 2; Page 6265, Algorithm 1, Column 2, Paragraph 4, “…Algorithm 1 shows the pseudo-code of the learning process … This objective penalizes the decrease of the discriminative power and forces the system to summarize a few generic decision modes for explanation …”. The examiner interprets the system/electronic device capable of implementing the code/methods/algorithms of learning decision trees; thus, the device contains a memory storing executable instructions; and at least one processor coupled to the memory, wherein when executing the executable instructions, the at least one processor is configured to implement the code/methods/algorithms of learning decision trees)); thus, Claims 15 and 16 are rejected for reasons set forth in the rejections of Claims 7 and 8, respectively.
Regarding Claim 19 and 20:
Claims 19 and 20 incorporate substantively all the limitations of Claims 7 and 8 in an non-transitory storage medium storing computer executable instructions and further recites wherein when the computer executable instructions are executed on a computer, the computer is triggered to: (Zhang, Page 6261, Figure 2; Page 6265, Algorithm 1, Column 2, Paragraph 4, “…Algorithm 1 shows the pseudo-code of the learning process … This objective penalizes the decrease of the discriminative power and forces the system to summarize a few generic decision modes for explanation …”. The examiner interprets the system/electronic device capable of implementing the code/methods/algorithms of learning decision trees; thus, the device contains The decision tree learning pipeline, trains, optimizes and evaluates on a computing device in which the CPU (processor), memory, and non-transitory storage medium are inherent)); thus, Claims 19 and 20 are rejected for reasons set forth in the rejections of Claims 7 and 8, respectively.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to IBRAHIM RAHMAN whose telephone number is (703)756-1646. The examiner can normally be reached M-F 8am-5pm.
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, Kakali Chaki can be reached at (571) 272-3719. 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.
/I.R./Examiner, Art Unit 2122
/KAKALI CHAKI/Supervisory Patent Examiner, Art Unit 2122