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 .
Detailed Action
The following action is in response to the communication(s) received on 8/11/2026.
As of the claims filed 8/11/2026:
Claims 1, 6, 8, 13, 14, and 15 have been amended.
Claims 3-5 and 10-12 have been canceled.
Claims 1, 6-8, and 13-17 are now pending.
Claims 1, 8, 14, and 15 are independent claims.
Continued Examination Under 37 CFR 1.114
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on 8/11/2026 has been entered.
Response to Arguments
Applicant’s arguments filed 8/11/2026 have been fully considered, but are not fully persuasive.
With respect to the interpretation 35 USC § 112(f):
The amendments have withdrawn the claim language which required this interpretation for the remaining limitation in claim 8 (“a first output data generation unit…”); thus, 35 USC § 112(f) has been withdrawn for claim 8.
With respect to the rejection under 35 USC § 101:
Applicant asserts that the claims provides its output to the target device comprising limited hardware resources, thus overcoming a technical challenge (p.13 ¶2). Examiner respectfully submits that the improvement is directed towards saving algorithmic complexity when adding new values to the decision tree. However, the method of performing such calculations does not require a particular architecture involved in training a decision tree (“additional learning processing” does not constitute additional training of a machine learning model). Thus, the amended claims involving "an embedded control device with limited hardware resources" are merely generally linked to the improved abstract idea.
Applicant further asserts that the claims eliminate structural changes to the decision tree as a whole, which guarantees time control cycle and reliability of the embedded hardware (p.14 ¶1). However, the present invention modifying a leaf node in a decision tree does not involve a particular training architecture of a machine learning model and thus does not require the hardware present to perform the improved method of updating the leaf node. Thus, the present hardware is merely generally linked as a particular application of the abstract idea.
Applicant further asserts that controlling the target device and outputting the target device with the output data defines the method of execution of the mathematical operations, thus reciting a technical solution to a technical problem (p.14 ¶2). Examiner respectfully submits that, as explained above, the target device is not explicitly required in the present claims, where the leaf nodes are updated through mere math operations instead of an architectural modification of a training technology of a machine learning model.
Applicant further asserts that the present invention is not well-understood, routine, or conventional to the prior art, and thus recites “significantly more” than the abstract idea under Step 2B of the patent eligibility analysis (p.15¶1). Examiner respectfully disagrees, as Step 2B considers whether the combinations of the claimed invention would amount to something "significantly more" than the abstract idea; however, as mentioned above, the crux of the claim corresponds to improving an algorithm through a mathematical operation. Thus, the significance is merely further directed towards an abstract idea.
Examiner respectfully submits that the improvement does not use the computer as merely a tool for performing the abstract ideas (p.15 ¶2). Examiner respectfully submits that, as explained above, the computer is not required to perform the abstract ideas as currently recited, thus the computer cannot be more than a mere application of the abstract ideas.
Thus, the claims remain ineligible.
With respect to the art rejection under 35 USC § 102:
Applicant asserts that Zhang does not teach a direct updating of device control output data using arithmetic or weighted average of pre-update output data and actual data (p.17¶2). Examiner respectfully submits that the argument has been carefully considered but is now moot further in view of the additional art from Xie: direct update of the leaf node without involving a change in structure of a whole learned decision tree or a change in a branch condition of the whole learned decision tree is further taught by Xie [p.10 ¶1] (“update the probabilities of the leaf node without structure modification of a decision tree”), where the new data set of class corresponds to the updated output data, and the updated probabilities of class k using eq. 10 corresponds to the weighted average.
Applicant further asserts that the prior art does not teach a learning rate when there are no changes in branch conditions of the whole learned decision tree (p.17 last ¶-p.18 ¶2). Examiner respectfully submits that the present invention does not recite a positive language of the learning rate when updating the leaf nodes. i.e., learning rate is not a required embodiment of the present invention. Thus, the claims remain unpatentable under Zhang/Xie.
Applicant further asserts that Zhang rebuilds the sub-tree for new data and thus does not disclose the amended claims involving no change in structure of the whole learned decision tree (p.18 ¶3-p.19¶1). Examiner respectfully submits that, as explained above, the argument is moot in view of the new prior art from Xie’s method of updating the leaf nodes from new data without structure modification of the decision tree.
Thus, Zhang/Xie remain teaching the present invention.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1, 6-8, and 13-17 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Claim 1 recites An embedded control device, thus a machine, one of the four statutory categories of patentable subject matter (Step 1). However, Claim 1 further recites:
identify an output node corresponding to the input data and generate associated output data...;, which are evaluations or judgements that can be performed in the human mind;
perform additional learning processing for the learned decision tree by generating updated output data by updating the output data associated with the output node, based on the output data and the actual data without involving a change in structure of the whole learned decision tree or a change in a branch condition of the whole learned decision tree, which are evaluations or judgements that can be performed in the human mind;
the updated output data is: an arithmetic average value of the output data before updating and the actual data, which is a mathematical concept;
a weighted average value of the output data before updating and the actual data, with reference to the number of data pieces associated with the output node, which is a mathematical concept;
adding, to the output data before updating, a result of multiplying a difference between the output data before updating and the actual data by a learning rate, which is a mathematical concept.
Thus, the claim recites an abstract idea under Step 2A Prong 1.
Under Step 2A Prong 2, the claim does not include any additional elements which integrate the abstract idea into a practical application, since the additional elements consist of:
A memory and a processor, the processor configured to…, as the performance of an abstract idea on a computer is not more than instructions to "apply it" on a computer, which by MPEP 2106.05(f) cannot integrate an abstract idea into a practical application;
acquire, as input data, data acquired from a target device, which is merely an insignificant extra-solution activity of data gathering, which by MPEP 2106.05(g) cannot integrate an abstract idea into a practical application;
provide the target device with the output data, which is merely an insignificant extra-solution activity of data transfer, which by MPEP 2106.05(g) cannot integrate an abstract idea into a practical application;
acquire actual data acquired from the target device, the actual data corresponding to the input data, which is merely an insignificant extra-solution activity of data gathering, which by MPEP 2106.05(g) cannot integrate an abstract idea into a practical application;
and store, in the memory, the learned decision tree for which the additional learning processing is performed, which is merely an insignificant extra-solution activity of data storage, which by MPEP 2106.05(g) cannot integrate an abstract idea into a practical application.
Thus, the claim is directed towards an abstract idea.
Further, the additional elements, alone or in combination, do not provide significantly more than the abstract idea itself, because implementation on a computer (MPEP 2106.05(f)) and the activity of data gathering (MPEP 2106.05(g)) cannot provide significantly more, as storing and retrieving information in memory is well understood, routine, and conventional (MPEP 2106.05(d)(II)(iv)); the activity of data transfer (MPEP 2106.05(g)) cannot provide significantly more, as receiving or transmitting data over a network is well understood, routine, and conventional (MPEP 2106.05(d)(II)(i), buySAFE, Inc. v. Google, Inc). The combination of these additional elements does not provide an inventive concept; thus, the claim remains ineligible. The combination of additional elements does not provide an inventive concept. Thus, the claim is ineligible.
Claim 6, dependent upon Claim 5, further recites
the learning rate changes according to the number of times the additional learning processing is performed, which is merely a detail of an abstract idea (multiplying a difference between the output data before updating and the actual data by a learning rate).
Thus, the claim recites an abstract idea under Step 2A Prong 1.
Under Step 2A Prong 2 and 2B, the claim does not recite any new additional elements which could integrate the abstract idea into a practical application or provide significantly more than the abstract idea itself. Thus, the claim is ineligible.
Claim 7, dependent upon Claim 1, further recites no additional abstract ideas. However:
Under Step 2A Prong 2, the claim does not include any additional elements which integrate the abstract idea into a practical application, since the additional elements consist of:
the learned decision tree is one of a plurality of decision trees for ensemble learning, which merely specifies the particular field of use or particular technological environment in which the abstract idea is to be performed, which by MPEP 2106.05(h) cannot integrate the abstract idea into a practical application.
Thus, the claim is directed towards an abstract idea.
Further, under Step 2B, the additional element does not provide significantly more than the abstract idea itself, because the particular field of use or particular technological environment (MPEP 2106.05(h)) cannot provide significantly more. Thus, the claim is ineligible.
Claim 8 recites an embedded control device, thus a machine, one of the four statutory categories of patentable subject matter (Step 1). However, Claim 8 further recites:
generate first output data, which is an evaluation or judgement that can be performed in the human mind;
generate second output data by identifying an output node corresponding to the reference input data , which is an evaluation or judgement that can be performed in the human mind;
and generated by performing machine learning based on the training input data and differential training data between the output data , which is an evaluation or judgement that can be performed in the human mind;
generate final output data..., which is an evaluation or judgement that can be performed in the human mind;
and perform additional learning processing for the learned decision tree by generating updated output data by updating the second output data associated with the output node, based on the second output data and differential data between the first output and the reference correct data without involving a change in structure of the whole learned decision tree or a change in a branch condition of the whole learned decision tree, which is an evaluation or judgement that can be performed in the human mind;
wherein the updated output data is: an arithmetic average value of the second output data before updating and the differential data, which is merely a detail of an abstract idea (performs additional learning processing);
the updated output data is a weighted average value of the second output data before updating and the differential data, with reference to the number of data pieces associated with the output node, which is a mathematical concept;
adding, to the second output data before updating, a result of multiplying a difference between the second output data before updating and the differential data by a learning rate, which is a mathematical concept.
Thus, the claim recites an abstract idea under Step 2A Prong 1.Under Step 2A Prong 2, the claim recites:
An embedded control device comprising: a memory and a processor, the processor configured to, as the performance of an abstract idea on a computer is not more than instructions to 'apply it' on a computer, which by MPEP 2106.05(f) cannot integrate an abstract idea into a practical application;
acquire reference input data from a target device, which is merely an insignificant extra-solution activity of data gathering, which by MPEP 2106.05(g) cannot integrate an abstract idea into a practical application;
inputting the reference input data into a model generated based on training input data and training correct data corresponding to the training input data, which is merely an insignificant extra-solution activity of data gathering, which by MPEP 2106.05(g) cannot integrate an abstract idea into a practical application;
by inputting the reference input data into a latest learned decision tree stored in the memory , which is merely an insignificant extra-solution activity of data gathering, which by MPEP 2106.05(g) cannot integrate an abstract idea into a practical application;
generated by inputting the training input data into the model and the training correct data , which is merely an insignificant extra-solution activity of data gathering, which by MPEP 2106.05(g) cannot integrate an abstract idea into a practical application;
to be used to control the target device, based on the first output data and the second output data; , which merely specifies the particular field of use or particular technological environment in which the abstract idea is to be performed, which by MPEP 2106.05(h) cannot integrate the abstract idea into a practical application;
provide the target device with the final output data which is used to control the target device, which is merely an insignificant extra-solution activity of data transfer, which by MPEP 2106.05(g) cannot integrate an abstract idea into a practical application;
acquire reference correct data from the target device, the correct data corresponding to the input data, which is merely an insignificant extra-solution activity of data transfer, which by MPEP 2106.05(g) cannot integrate an abstract idea into a practical application;
and store, in the memory, the learned decision tree for which the additional learning processing is performed, which is merely an insignificant extra-solution activity of data transfer, which by MPEP 2106.05(g) cannot integrate an abstract idea into a practical application.
Thus, the claim is directed towards and abstract idea.
Further, the additional element(s), alone or in combination, do not provide significantly more than the abstract idea itself, because the activity of data gathering (MPEP 2106.05(g)) cannot provide significantly more, as storing and retrieving information in memory is well understood, routine, and conventional (MPEP 2106.05(d)(II)(iv)); the activity of data transfer (MPEP 2106.05(g)) cannot provide significantly more, as receiving or transmitting data over a network is well understood, routine, and conventional (MPEP 2106.05(d)(II)(i), buySAFE, Inc. v. Google, Inc); the particular field of use or particular technological environment (MPEP 2106.05(h)) cannot provide significantly more; implementation on a computer (MPEP 2106.05(f)) cannot provide significantly more. The combination of these additional elements does not provide an inventive concept; thus, the claim remains ineligible.
Claim 13, dependent upon Claim 8, further recites
the learning rate changes according to the number of times the additional learning processing is performed, which is merely a detail of a judicial exception (multiplying a difference between the output data before updating and the actual data by a learning rate).
Thus, the claim recites an abstract idea under Step 2A Prong 1.
Under Step 2A Prong 2 and 2B, the claim does not recite any new additional elements which could integrate the abstract idea into a practical application or provide significantly more than the abstract idea itself. Thus, the claim is ineligible.
Claim 14 recites A learning method, thus a process, one of the four statutory categories of patentable subject matter. However, Claim 14 recites precisely the abstract ideas and additional elements of Claim 1. Therefore, Step 2A Prong 1, Step 2A Prong 2, and Step 2B analyses remain the same. Claim 14 is rejected as subject-matter ineligible for reasons set forth in the rejections of Claim 1.
Claim 15 recites A control method, thus a process, one of the four statutory categories of patentable subject matter. However, Claim 15 recites precisely the abstract ideas and additional elements of Claim 8. Therefore, Step 2A Prong 1, Step 2A Prong 2, and Step 2B analyses remain the same. Claim 15 is rejected as subject-matter ineligible for reasons set forth in the rejections of Claim 8.
Claim 16, dependent upon Claim 14, further recites no additional abstract ideas. However:
Under Step 2A Prong 2, the claim does not include any additional elements which integrate the abstract idea into a practical application, since the additional elements consist of:
A non-transitory computer readable medium storing a control program for a device for executing the control method, as the performance of an abstract idea on a computer is not more than instructions to "apply it" on a computer, which by MPEP 2106.05(f) cannot integrate an abstract idea into a practical application.
Thus, the claim is directed towards an abstract idea.
Further, under Step 2B, the additional element does not provide significantly more than the abstract idea itself, because implementation on a computer (MPEP 2106.05(f)) cannot provide significantly more. Thus, the claim is ineligible.
Claim 17, dependent upon Claim 15, further recites no additional abstract ideas. However:
Under Step 2A Prong 2, the claim does not include any additional elements which integrate the abstract idea into a practical application, since the additional elements consist of:
A non-transitory computer readable medium storing a control program for a device for executing the control method, as the performance of an abstract idea on a computer is not more than instructions to "apply it" on a computer, which by MPEP 2106.05(f) cannot integrate an abstract idea into a practical application.
Thus, the claim is directed towards an abstract idea.
Further, under Step 2B, the additional element does not provide significantly more than the abstract idea itself, because implementation on a computer (MPEP 2106.05(f)) cannot provide significantly more. Thus, the claim is ineligible.
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.
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, 6-8, and 13-17 are rejected under 35 U.S.C. 103 as being unpatentable over Zhang et al., "Persistent Anytime Learning of Objects from Unseen Classes" (hereinafter Zhang) further in view of Xie et al., "hi-RF: Incremental Learning Random Forest for large-scale multi-class Data Classification" (hereinafter Xie).
Regarding Claim 1, Zhang teaches:
An embedded control device comprising: a memory and a processor, the processor configured to:
(Zhang [p.15 2nd ¶] iGBDT and GBDT are both implemented using the Java programming language.) (Note: the original data corresponds to the reference input data; using the Java programming language requires a memory and a processor.)
acquire, as input data, data acquired from a target device; (Zhang [p.11 ¶4] iGBDT merges the original data D1 and the new batch data D2 (both ordered according to attribute k) to obtain an ordered data set D3. So the number of samples in D3 = D2 + D1
[p.15 2nd ¶] iGBDT and GBDT are both implemented using the Java programming language.) (Note: the original data corresponds to the reference input data; using the Java programming language requires memory, which corresponds to the target device.)
identify an output node corresponding to the input data and generate associated output data to be used to control the target device, through inference processing using a latest learned decision tree stored in the memory; provide the target device with the output data;
(Zhang, Algorithm 2,
PNG
media_image1.png
153
522
media_image1.png
Greyscale
[p.6 ¶3] In line 1, GBDT first initialises the model with the mean value of the class values of all the instances.
[p.9 3rd ¶] In the process of building the GBDT model, the key point is to find the best split attribute which makes sure the square loss error is minimised.
[p.10 3rd ¶] After the best split points for all the attributes have been computed, the attribute that has the smallest sum least square loss error will be chosen as the best split attribute, the corresponding best split point (value) will be used to horizontally split the data into the left and right parts. Then the process for finding the best split attribute (and point) for a given node in GBDT finishes.
[p.11 section 4.3] As new batches of data are continuously arriving, the ensemble model built by GBDT which may contain thousands of decision trees should be timely updated so as to accurately model/reflect the complete data which contains the newly arrived instances) (Note: x corresponds to the input data; class value yi is related to x, thus based on the training input data; F0(x) corresponds to the generated first output data; the class values (yi) correspond to the training correct data; the given node when the GBDT finishes corresponds to the output node; the best split attribute corresponds to the output data associated with the output node.)
acquire actual data acquired from the target device, the actual data corresponding to the input data;
(Zhang [p.11 ¶4] iGBDT merges the original data D1 and the new batch data D2 (both ordered according to attribute k) to obtain an ordered data set D3. So the number of samples in D3 = D2 + D1
) (Note: the new batch data correspond to the actual data)
Zhang does not teach, but Xie further teaches:
perform additional learning processing for the learned decision tree by generating updated output data by updating the output data associated with the output node, based on the output data and the actual data without involving
(Xie [p.10 ¶1] ReGenerate leaves probabilities (RLP) is a light weight method to update the probabilities of the leaf node without structure modification of a decision tree, which contributes a lot to computational cost reduction. The whole procedure is described in Algorithm 4 and will be explained in detail as following.
PNG
media_image2.png
687
424
media_image2.png
Greyscale
RLP also takes use of the threshold generated by OOB estimation. For each tree in RF, if the out-of-bag error is more than threshold means the versions of trees reach the average level and are robust in learning information. Thus, RLP is to update these trees. Usually, once a decision tree is constructed, the split function in each intern node is determined and the probabilities of leaves is generated. RLP is proposed to update that. Since the splitting function is unchanged, we can put the new data into the tree Ti and get a new data set of class in each leaf node j, whose number for a tree is n node.) (Note: the claims recite two embodiments, where it requires either no change in structure or no change in branch condition; Xie teaches the embodiment of without involving… a change in a branch condition of the whole learned decision tree; the new data set of class corresponds to the updated output data; the resulting trees (in line 21) correspond to each of the stored learned decision trees)
wherein the updated output data is: an arithmetic average value of the output data before updating and the actual data; a weighted average value of the output data before updating and the actual data, with reference to the number of data pieces associated with the output node; or a value obtained by adding, to the output data before updating, a result of multiplying a difference between the output data before updating and the actual data by a learning rate.
(Xie [p.10, eq.1 line 13]
PNG
media_image3.png
693
429
media_image3.png
Greyscale
[p.11 eq.10]
PNG
media_image4.png
129
706
media_image4.png
Greyscale
) (Note: the updated probabilities are reweighted based on the new data within the all training data D; thus, the updated probabilities correspond to the weighted average value of the output data before updating and the actual (new) data)
Xie and Zhang are analogous to the present invention because both are from the same field of endeavor of efficient updating methods of decision trees. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to implement RLP from Xie into Zhang’s ensemble model updating method. The motivation would be to “ReGenerate leaves probabilities (RLP) is a light weight method to update the probabilities of the leaf node without structure modification of a decision tree, which contributes a lot to computational cost reduction.” (Xie p.10 ¶1).
Regarding Claim 6, since the claim language is part of an optional embodiment (of using a learning rate), this claim remains unpatentable under Zhang/Xie.
Regarding Claim 7, Zhang/Xie respectively teaches and incorporates the claimed limitations and rejections of Claim 1. Xie, via Zhang/Xie, further teaches:
The control device according to claim1, wherein the learned decision tree is one of a plurality of decision trees for ensemble learning.
(Xie [p.10 ¶2] For each tree in RF, if the out-of-bag error is more than threshold means the versions of trees reach the average level and are robust in learning information. Thus, RLP is to update these trees.)
Regarding Claim 8, Zhang teaches:
An embedded control device comprising: a memory and a processor, the processor configured to:
(Zhang [p.15 2nd ¶] iGBDT and GBDT are both implemented using the Java programming language.) (Note: the original data corresponds to the reference input data; using the Java programming language requires a memory and a processor.)
acquire reference input data from a target device;
(Zhang [p.11 ¶4] iGBDT merges the original data D1 and the new batch data D2 (both ordered according to attribute k) to obtain an ordered data set D3. So the number of samples in D3 = D2 + D1
[p.15 2nd ¶] iGBDT and GBDT are both implemented using the Java programming language.) (Note: the original data corresponds to the reference input data; using the Java programming language requires memory, which corresponds to the target device.)
generate first output data by inputting the reference input data into a model generated based on training input data and training correct data corresponding to the training input data;
(Zhang, Algorithm 2,
PNG
media_image1.png
153
522
media_image1.png
Greyscale
[p.6 ¶3] In line 1, GBDT first initialises the model with the mean value of the class values of all the instances.
[p.9 3rd ¶] In the process of building the GBDT model, the key point is to find the best split attribute which makes sure the square loss error is minimised.) (Note: x corresponds to the training input data; class value yi is related to x, thus based on the training input data; F0(x) corresponds to the generated first output data; the class values (yi) correspond to the training correct data; each split attribute output corresponds to each output hereon)
generate second output data by identifying an output node corresponding to the reference input data by inputting the reference input data into a latest learned decision tree stored in the memory and generated by performing machine learning based on the training input data and differential training data between the output data generated by inputting the training input data into the model and the training correct data;
(Zhang, Algorithm 2,
PNG
media_image1.png
153
522
media_image1.png
Greyscale
[p.6 ¶3] In line 1, GBDT first initialises the model with the mean value of the class values of all the instances. Then M = pt models (trees) of depth ≤ pd are sequentially learned in M steps using the for loop. In line 3, under the least square loss function, GBDT computes the residual of the existing m − 1 models, which have already been learned so far. With the above residual as the class values of the corresponding instances in the training data, it trains the mth decision tree model using CART. Since this a case of tree models, a different multiplier γm is computed for every leaf. Then, by incorporating the new decision tree model, it updates the ensemble model Fm−1(x) to form the new ensemble model Fm(x), through linear superposition (line 5).) (Note: each iteration of the previous and updated Fm(x) corresponds to the learned decision tree; each updated split attribute of the updated learned decision tree corresponds to the second output data; each residual (line 3) corresponds to each differential training data between each output data and the training correct data)
generate final output data to be used to control the target device, based on the first output data and the second output data; provide the target device with the final output data which is used to control the target device; acquire reference correct data from the target device, the correct data corresponding to the input data;
(Zhang [p.11 ¶4] iGBDT merges the original data D1 and the new batch data D2 (both ordered according to attribute k) to obtain an ordered data set D3. So the number of samples in D3 = D2 + D1
[p.11 section 4.3] As new batches of data are continuously arriving, the ensemble model built by GBDT which may contain thousands of decision trees should be timely updated so as to accurately model/reflect the complete data which contains the newly arrived instances
[p.15 2nd ¶] iGBDT and GBDT are both implemented using the Java programming language.) (Note: the new batch data correspond to the reference correct data; the timely updated ensemble model using the new batches of data corresponds to the final output data; iGBDT receiving the new batches of data corresponds to the target device being provided the final output data)
Zhang does not teach, but Xie further teaches:
and perform additional learning processing for the learned decision tree by generating updated output data by updating the second output data associated with the output node, based on the second output data and differential data between the first output data and the reference correct data without involving a change in structure of the whole learned decision tree or a change in a branch condition of the whole learned decision tree, and store, in the memory, the learned decision tree for which the additional learning processing is performed,
(Xie [] (Xie [p.10 ¶1] ReGenerate leaves probabilities (RLP) is a light weight method to update the probabilities of the leaf node without structure modification of a decision tree, which contributes a lot to computational cost reduction. The whole procedure is described in Algorithm 4 and will be explained in detail as following.
PNG
media_image2.png
687
424
media_image2.png
Greyscale
RLP also takes use of the threshold generated by OOB estimation. For each tree in RF, if the out-of-bag error is more than threshold means the versions of trees reach the average level and are robust in learning information. Thus, RLP is to update these trees. Usually, once a decision tree is constructed, the split function in each intern node is determined and the probabilities of leaves is generated. RLP is proposed to update that. Since the splitting function is unchanged, we can put the new data into the tree Ti and get a new data set of class in each leaf node j, whose number for a tree is n node.) (Note: the new data set of class corresponds to the updated output data; the resulting trees (in line 21) correspond to each of the stored learned decision trees)
wherein the updated output data is: an arithmetic average value of the second output data before updating and the differential data; a weighted average value of the second output data before updating and the differential data, with reference to a number of data pieces associated with the output node; or a value obtained by adding, to the second output data before updating, a result of multiplying a difference between the second output data before updating and the differential data by a learning rate.
(Xie [p.10, eq.1 line 13]
PNG
media_image3.png
693
429
media_image3.png
Greyscale
[p.11 eq.10]
PNG
media_image4.png
129
706
media_image4.png
Greyscale
) (Note: the updated probabilities are reweighted based on the new data within the all training data D; thus, the updated probabilities correspond to the weighted average value of the output data before updating and the actual (new) data)
Xie and Zhang are analogous to the present invention because both are from the same field of endeavor of efficient updating methods of decision trees. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to implement RLP from Xie into Zhang’s ensemble model updating method. The motivation would be to “ReGenerate leaves probabilities (RLP) is a light weight method to update the probabilities of the leaf node without structure modification of a decision tree, which contributes a lot to computational cost reduction.” (Xie p.10 ¶1).
Regarding Claim 13, since the claim language is part of an optional embodiment (of using a learning rate), this claim remains unpatentable under Zhang/Xie.
Independent Claim 14 recites a method to perform precisely the methods of Claim 1. Thus, Claim 14 is rejected for reasons set forth in Claim 1.
Independent Claim 15 recites A control method to perform precisely the methods of Claim 8. Thus, Claim 15 is rejected for reasons set forth in Claim 1.
Claim 16, dependent on claim 14, recites A non-transitory computer readable medium storing a control program for a device for executing (Zhang [p.15 2nd ¶] iGBDT and GBDT are both implemented using the Java programming language.) (Note: the original data corresponds to the reference input data; using the Java programming language requires non-transitory computer-readable medium.) precisely the methods of Claim 1. Thus, Claim 16 is rejected for reasons set forth in Claim 1.
Claim 17, dependent on claim 15, recites A non-transitory computer readable medium storing a control program for a device for executing (Zhang [p.15 2nd ¶] iGBDT and GBDT are both implemented using the Java programming language.) (Note: the original data corresponds to the reference input data; using the Java programming language requires a non-transitory computer-readable medium.) precisely the methods of Claim 8. Thus, Claim 17 is rejected for reasons set forth in Claim 8.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JOSEP HAN whose telephone number is (703)756-1346. The examiner can normally be reached Mon-Fri 9am-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 on (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.
/J.H./Examiner, Art Unit 2122
/KAKALI CHAKI/Supervisory Patent Examiner, Art Unit 2122
Prosecution Insights