DETAILED ACTION
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 .
Response to Amendments
This action is in response to the submission filed 26 May 2026 for application 18/024,903. Currently claims 1-18 are canceled. Claim 27 has been amended. Claims 19-38 are pending and have been examined.
The objection to the drawings are withdrawn in view of the new drawings submitted.
The objection to claim 27 has been withdrawn in view of the amendments made.
The §112(b) rejection of claims 22 and 34 has been withdrawn in view of the arguments presented.
Response to Arguments
Regarding applicant’s arguments, filed 26 May 2026, in regards to the rejection of claims under 35 U.S.C. §101, see page 8, Applicant argues that Claims 19-38 are rejected as allegedly being directed to ineligible subject matter, on the grounds that the claims are directed to "abstract ideas without significantly more." The Applicant respectfully disagrees. While the claims undoubtedly include mathematical operations, which standing alone could be considered abstract ideas, each claim as a whole is properly directed to an improvement to a field of technology. Specifically, the claims are directed to improved techniques for selecting from among a plurality of machine-learning source domain candidates, so as to apply the best one in a target domain having a new or changed execution environment.
Examiners response: Applicant’s arguments have been fully considered but they are not persuasive. Examiner respectfully disagrees that claims are patentable because although Applicant argues that claims are directed to an improvement to a field of technology, Applicant themselves immediately state that the claims are directed to improved techniques for “selecting from among a plurality of machine-learning source domain candidates” which is identified as an abstract idea in Step 2A, prong 1 of the 101 analysis. Here, the improvement is in the abstract idea of “selecting”. As disclosed in MPEP 2106.05(a) it is important to note that the judicial exception alone cannot provide the improvement. Hence, the rejection is maintained.
Regarding applicant’s arguments, filed 26 May 2026, in regards to the rejection of claims under 35 U.S.C. §101, see pages 8 and 9, Applicant argues that the Office Action's analysis of claim 19, at pages 4 and 5 of the action, follows the framework set forth in MPEP 2106 but goes awry at a few spots. For example, after correctly noting that each of the claims is directed to one of the four statutory categories, the Office Action performs "Step 2A, prong 1" of the analysis, concluding, without any substantial discussion of the claim language, that all the steps but the last are mental processes that "could be performed in the mind or with the aid of pencil and paper..." In fact, humans cannot practically carry out the recited steps in their minds.
For example, humans do not and cannot practically calculate, for each of multiple machine-learning source domain candidates, a diversity metric representing a marginalized measure of sample diversity. Even if a human could readily access the data samples for each machine-learning model, which is not a given, the calculations for a marginalized measure of diversity are such that even in a computing device these measures are typically computed using numerical methods. (See Specification p. 7, line 14 - p. 8, line 3.) They simply are not the sort of calculations that can reasonably be considered a "mental process," and the person of ordinary skill in the art would immediately appreciate that this step, at least, is not a "mental process" or "abstract idea."
Examiners response: Applicant’s arguments have been fully considered but they are not persuasive. Examiner respectfully disagrees that claims are patentable because firstly, Applicant does not explain why the calculation cannot be performed by humans even if they are typically performed computed using numerical methods. Just because something is typically calculated using a machine does not mean that it cannot be calculated by a human either mentally or with the aid of a paper/pen. Secondly, in the instant case the identifying and selecting limitations can be easily done by a human observing the candidates and making an evaluation of the diversity metric and selecting the highest. Even the calculating limitation is recited without any details and is so broad that a human can easily evaluate the diversity in the samples of the source domain candidates. Hence, the claims fall under the “mental process” grouping of abstract ideas.
Regarding applicant’s arguments, filed 26 May 2026, in regards to the rejection of claims under 35 U.S.C. §101, see pages 9-11, Applicant argues that in addition, the Office Action goes on to assert that the remaining step of claim 19, "applying the selected machine-learning source domain candidate to a target domain in a new or changed execution environment" is "recited so generically" that it represents no more than mere instructions to apply the judicial exception on a computer. The analysis continues that "mere instructions to apply an exception cannot provide an inventive concept."
This is incorrect, in several ways. First, the supposed "judicial exception" here is the alleged "mental processes" covered by the preceding three steps of the claim. This last step, "applying the selected machine-learning source domain candidate to a target domain" is not simply a step of applying/performing those first three steps on a computer. Instead, this last step specifies that the selected one of the plurality of machine-learning source domain candidates is applied to a target domain having a new or changed execution environment. Thus, the result of the first three steps is put to work in a particular environment.
This is not, by the way, an "insignificant post-solution activity." Instead, this is the realization of the solution itself, i.e., the selection and application of the best one of the multiple machine-learning source domain candidates to a target domain having a new or changed execution environment.
The Office Action's analysis does not acknowledge or identify the "inventive concept" of the claim, which can be characterized as choosing and applying a machine-learning source domain candidate to a target domain in a new/changed execution environment based on diversity metrics computed for the multiple possible candidates. Note that this inventive concept is not seeing to "tie up" the "diversity metric" itself, or the calculation of diversity metrics, but is putting these diversity metrics to use in a practical way, having real-world effects.
Indeed, this is the improvement to a technical field that the present claims are directed to. Given a plurality of available machine-learning source domain candidates that can possibly be applied to a target domain in a new or changed execution environment, the claims provide a structured technique, based on these real-world machine-learning source domain candidates, for selecting and applying the most suitable one for a target domain in a new or changed execution environment. This improvement is what transforms the underlying mathematical calculations, which cannot be practically carried out in a human mind in any event, into a practical application of those calculations.
Note that the Desjardins decision emphasized the purported "improvement," noting that the application at issue identified an improvement addressed by the claims and that this improvement was reflected in the claims. (Desjardins pp. 8, 9.) The same is true here. The present Specification, beginning at page 4, line 10, discusses problems with conventional approaches for source domain selection when there is limited information about the target domain, which makes transfer learning (where a machine-learning model trained on a source domain is deployed in a different target domain) challenging. The Specification explains that the presently claimed techniques address this problem by focusing on diversity in the machine- learning source domain candidates. (Specification p. 4 1. 10 - p 5. 1. 37.) Even a cursory review of the claims indicates that this improvement is directly reflected in the claims, which are directed to choosing the appropriate machine-learning source domain candidate for deployment based on metrics of diversity computed for several candidates.
Thus, claim 19 and the other pending claims are all directed to an improvement in a field of technology, specifically to the field of transfer learning using machine-learning models. The rejections under 35 U.S.C. § 101 should be withdrawn.
Examiners response: Applicant’s arguments have been fully considered but they are not persuasive. Examiner respectfully disagrees that claims are patentable because in Step 2A, prong 2 of the 101 analysis the “applying” limitation is an additional element with no details of what exactly applying the selected machine learning source domain candidate to a target domain entails. There are no details of the actual “applying” step recited in the claim. The claim merely recites applying the selected source domain to a target domain in a new environment and this is clearly represents no more than mere instructions to apply the judicial exceptions (identified in Step 2A, prong 1) on a computer. The office action does not state that this is an insignificant post-solution activity. Again, Applicant states on Page 9 (last paragraph) that the “inventive concept” of the claim can be characterized as choosing a machine learning source domain candidate, which as explained above is an abstract idea. Also, the instant case is different from Desjardins because in the instant case the improvement is in choosing the machine learning source domain candidate based on the diversity metric, which is an abstract idea. As disclosed in MPEP 2106.05(a) it is important to note that the judicial exception alone cannot provide the improvement. Lastly, although applicant argues that claims are directed to improvement to the field of transfer learning it is only recited in claim 24 (and similarly in claim 36) and again with no details of the actual transfer learning other than that the applying comprises of transfer learning. Hence, the rejection is maintained.
Regarding applicant’s arguments, filed 26 May 2026, in regards to the rejection of claims under 35 U.S.C. §103, see pages 12 and 13, Applicant argues that the claims are thus directed to selecting a machine-learning source domain candidate, from several possible candidates, based on a diversity metric. Most of the Office Action's analysis of claim 19 is based on the "Wu" reference, which is titled "Entropy Minimization vs. Diversity Maximization for Domain Adaptation." This reference certainly mentions "diversity." However, this reference describes techniques for adapting a machine-learning model to a particular pair of source domain and target domain, and does not involve the evaluation of and selection from a plurality of source domains.
The Office Action suggests, at page 19, that the "batch of source samples" shown in Wu's Figure 1 correspond to "source domain candidates." This is simply incorrect. Figure 1 illustrates a "batch of source samples" for a single "Source Domain," as well as a "batch of target samples" for a single "Target Domain." A "source sample" is not a "machine-learning source domain candidate," it is instead simply a "sample," i.e., an input used for training the machine- learning model in that domain. Indeed, it would make no sense for a "sample" to be a "machine- learning source domain candidate" according to the claim, because claim 19 specifies that a diversity metric is calculated for each of the identified plurality of machine-learning source domain candidates - one cannot calculate a diversity metric for a single sample.
Thus, Wu does not disclose either "identifying a plurality of machine-learning source domain candidates" or "calculating, for each of the identified machine-learning source domain candidates, a diversity metric ..." Wu instead is concerned with a single source domain. Regarding the step of "calculating ... a diversity metric" for each of the candidates, the Office Action points to Wu's Section III, which refers to a measure of "category diversity," but Wu does not include any suggestion that this is measured or calculated for each of a plurality of machine-learning source domain candidates. In fact, the "category diversity" discussed by Wu is not a measure of source domain diversity (or entropy), but is instead a measure of the diversity of the target domain samples.
Likewise, Wu does not disclose selecting an identified machine-learning source domain candidate from a plurality of such candidates based on calculated diversity metrics for the candidates. Instead, Wu is concerned with how to perform training, in a given source domain, to produce a machine-learning model having optimized diversity for a given target domain.
Thus, rather than disclosing steps for evaluating and selecting from among a plurality of machine-learning source domain candidates, based on a diversity metric calculated for each one, Wu teaches techniques for adapting a single machine-learning model, in a source domain, to better fit a target domain application of that model.
Examiners response: Applicant’s arguments have been fully considered but they are not persuasive. Examiner respectfully disagrees that reference Wu does not teach these limitations because, Page 1, Column 2, Last 2 Paragraphs of reference Wu states that 1) We propose a minimal-entropy diversity maximization (MEDM) method for UDA. 2) MEDM outperforms state-of-the-art methods on four domain adaptation datasets, including VisDA-2017, ImageCLEF, Office-Home and Office-31. This shows that there are a plurality of domain datasets being identified and from which the samples are taken. Also, Page 7, Column 1, Paragraph 1 states that Office-Home [44] is a typical dataset with a large number of classes (65 classes), which containing 15,500 images from four visually very different domains: Artistic images, Clip Art, Product images, and Real-world images. This again shows the diversity within the domain datasets. Hence, reference Wu teaches identifying a plurality of machine-learning source domain candidates; calculating, for each of the identified machine-learning source domain candidates, a diversity metric, the diversity metric representing a marginalized measure of sample diversity of the respective machine-learning source domain candidate.
Regarding applicant’s arguments, filed 26 May 2026, in regards to the rejection of claims under 35 U.S.C. §103, see pages 13 and 14, Applicant argues that the Office Action turns to the secondary reference, "Taylor," for teaching the step of "applying the selected machine-learning source domain candidate to a target domain in a new or changed execution device." Taylor is concerned generally with transfer learning, where a model trained in one domain (a source domain) is applied in another (a target domain), but so is Wu.
Taylor does not cure the deficiencies of Wu, because Taylor also fails to disclose or suggest that one of a plurality of machine-learning source domain candidates is selected based on diversity metrics calculated for each candidate. Indeed, it does not appear that Taylor mentions diversity at all.
Accordingly, the combination of Taylor with Wu fails to disclose or suggest the invention as claimed in claim 19. The rejection of claim 19 should be withdrawn, as should the rejections of its several dependent claims.
Examiners response: Applicant’s arguments have been fully considered but they are not persuasive. Examiner respectfully disagrees because reference Taylor is only relied upon to teach transfer in a new or changed environment and reference Wu is relied upon to teach the selecting of the source domain candidate from a plurality of machine-learning source domain candidates based on diversity metrics calculated for each candidate has explained above and shown in the detailed rejection below. Hence, the combination of Taylor with Wu teaches claim 19.
Regarding applicant’s arguments, filed 26 May 2026, in regards to the rejection of claims under 35 U.S.C. §103, see page 14, Applicant argues that the independent claim 31 is directed to a "server node" and is an apparatus counterpart to method claim 19, reciting operations that correspond directly to the steps of claim 19. Independent claim 38 is directed to a non-transitory computer-readable medium, again reciting operations directly corresponding to the steps of claim 19.
These claims are rejected over Wu and Taylor, further in view of Tan, where Tan is offered only for its teachings regarding components of a server that implements a "machine learning unit," where those components include a memory. The rejections of claims 31 and 38 are in error for the same reasons given above for claim 19, as these rejections expressly rely on the very same findings.
Examiners response: Applicant’s arguments have been fully considered but they are not persuasive. Examiner respectfully disagrees because, independent claims 31 and 38 are substantially similar to independent claim 19 and are therefore rejected for the same reasons as explained above and shown in the detailed rejection below.
Regarding applicant’s arguments, filed 26 May 2026, in regards to the rejection of claim 23 under 35 U.S.C. §103, see page 15, Applicant argues that at least some of these rejections are in error for additional reasons. For example, claim 23 specifies that the selecting step of claim 19 comprises selecting a plurality of machine- learning source domain candidates having diversity metrics above a threshold, and applying each of them to the target domain. The Office Action says that the selecting step here is disclosed in Tan, citing Tan's paragraphs 0032 and 0038, but this says nothing at all about selecting multiple machine-learning domain candidates. Instead, these paragraphs discuss badge readers that read badges worn by personnel entering a particular area, in the context of using a "local machine learning model" to predict the identity of a person based on images captured of that person, where the predicted identity is referred to as a "predicted label" or, simply, "label." (Tan 1 0029-0032.) Tan goes on to discuss handling of "data-label" pairs, which are simply the predicted identities provided together with the respective images. None of this has anything at all to do with selecting a plurality of machine-learning source domain candidates based on diversity metrics.
Examiners response: Applicant’s arguments have been fully considered but they are not persuasive. Examiner respectfully disagrees that reference Tan does not teach this limitation because paragraph [0038] states that the sampler 224 forwards correctly predicted data-label pairs where the output entropy value H exceeds a predetermined threshold and that other than entropy, the forwarding decision can also be made based on alternative functions of the probability distribution, such as diversity index. In machine learning, a source domain is the original dataset or distribution from which a model initially learns patterns and features. Source domain candidates refer to the pool of pre-existing, often labeled datasets considered for transfer learning or domain adaptation, helping a model succeed in a new, unlabeled target domain. So reference Tan forwarding correctly predicted data-label pairs corresponds to selecting a plurality of machine-learning source domain candidates and exceeding a predetermined threshold and that other than entropy, the forwarding decision can also be made based on alternative functions of the probability distribution, such as diversity index corresponds to having respective diversity metrics above a predetermined threshold. Hence, reference Tan teaches that limitation.
Regarding applicant’s arguments, filed 26 May 2026, in regards to the rejection of claim 24 under 35 U.S.C. §103, see pages 15 and 16, Applicant argues that regarding claim 24, the Office Action quotes a section of Taylor that describes transfer learning, noting that "how to best generate ... source tasks so that they are most likely to be useful for an arbitrary target task in the same domain in the same domain is an important area of open research." The Office Action includes a "note" that "Novel target task corresponds to detecting a change in the execution environment of the target domain," but of course this is incorrect. The reference is concerned with optimizing training in the source domain so that the trained machine-learning is best able to deal with any arbitrary target task, which of course may be a "novel" task. Nothing here suggests that any steps are carried out in response to detecting a novel task, nor is there anything that would suggest how this detecting would be done. The rejections of claims 24-27 should be withdrawn for this additional reason.
Examiners response: Applicant’s arguments have been fully considered but they are not persuasive. Examiner respectfully disagrees that the rejection of claim 24 should be withdrawn because detecting a change in the execution environment of the target domain is very broad claim language and hence under the broadest reasonable interpretation novel target task from reference Taylor corresponds to a change in the execution environment of the target domain because when something is novel or new it shows that there is a change in the environment. Furthermore, although applicant argues that nothing here suggests that any steps are carried out in response to detecting a novel task, nor is there anything that would suggest how this detecting would be done, Examiner disagrees because firstly, reference Taylor states on Page 1655 (Last but one Paragraph) that the goal of an agent is to perform as well as possible in a novel target task showing that it is performing some steps as well as possible in the novel task. Secondly, claim 24 does not recite any details of how the detecting is done and is broad. Hence, reference Taylor teaches that limitation.
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 19 - 38 are rejected under 35 U.S.C. 101 because the claimed invention is directed towards abstract ideas without significantly more.
Regarding claims 19-30:
According to the first step (Step 1) of the 101 analysis, claims 19-30 are directed to a method (process) and falls within one of the four statutory categories (i.e., process, machine, manufacture, or composition of matter).
Regarding claim 19:
In the next step (Step 2A, prong 1) of the analysis, the limitations of:
identifying a plurality of machine-learning source domain candidates;
calculating, for each of the identified machine-learning source domain candidates, a diversity metric, the diversity metric representing a marginalized measure of sample diversity of the respective machine-learning source domain candidate;
selecting the identified machine-learning source domain candidate having a highest diversity metric among the calculated diversity metrics;
Under the broadest reasonable interpretation, the above limitations are process steps that cover mental processes including an observation, evaluation, judgment or opinion that could be performed in the mind or with the aid of pencil and paper but for the recitation of a generic computer component. If a claim, under its broadest reasonable interpretation, covers a mental process but for the recitation of generic computer components, then it falls within the “Mental Process” grouping of abstract ideas.
In the next step (Step 2A, prong 2) of the analysis, the limitation
applying the selected machine-learning source domain candidate to a target domain in a new or changed execution environment.
is considered to be an additional element and it does not integrate the abstract idea into a practical application because the additional element is recited so generically (no details whatsoever are provided other than that it is a method that applies the selected machine-learning source domain candidate to a target domain in a new or changed execution environment) that it represents no more than mere instructions to apply the judicial exception on a computer. As discussed in MPEP 2106.05(f), mere instructions to implement an abstract idea on a computer as a tool to perform an abstract idea is not indicative of integration into a practical application.
In the last step (Step 2B) of the analysis, the additional element does not amount to significantly more than the judicial exceptions. As explained with respect to Step 2A Prong Two, the method that applies the selected machine-learning source domain candidate to a target domain in a new or changed execution environment, is at best the equivalent of merely adding the words “apply it” to the judicial exception. See MPEP 2106.05(f). Mere instructions to apply an exception cannot provide an inventive concept and does not amount to significantly more than the judicial exception. The claim is not patent eligible.
Regarding claim 20:
In the next step (Step 2A, prong 1) of the analysis, the limitation of:
wherein the diversity metric is calculated based on information theoretic measures.
Under the broadest reasonable interpretation, the above limitations are process steps that cover mental processes including an observation, evaluation, judgment or opinion that could be performed in the mind or with the aid of pencil and paper but for the recitation of a generic computer component. If a claim, under its broadest reasonable interpretation, covers a mental process but for the recitation of generic computer components, then it falls within the “Mental Process” grouping of abstract ideas.
In the next step (Step 2A, prong 2) of the analysis, it does not integrate into a practical application because it does not add any additional elements that integrate the abstract idea into practical application.
In the last step (Step 2B) of the analysis, it does not add any additional elements that amount to significantly more than the abstract idea and thus fails to add an inventive concept. The claim is not patent eligible.
Regarding claim 21:
In the next step (Step 2A, prong 1) of the analysis, the limitation of:
wherein the diversity metric is calculated based on a one- parameter measure of generalized entropy.
Under the broadest reasonable interpretation, the above limitations are process steps that recite mathematical relationships and calculations but for the recitation of generic computer components. If a claim, under its broadest reasonable interpretation covers mathematical concepts but for the recitation of generic computer components, then it falls within the “Mathematical concepts” grouping of abstract ideas.
In the next step (Step 2A, prong 2) of the analysis, it does not integrate into a practical application because it does not add any additional elements that integrate the abstract idea into practical application.
In the last step (Step 2B) of the analysis, it does not add any additional elements that amount to significantly more than the abstract idea and thus fails to add an inventive concept. The claim is not patent eligible.
Regarding claim 22:
In the next step (Step 2A, prong 2) of the analysis, the limitation
wherein the one-parameter measure is selected from the following: the Renyi entropy; the Havrda-Charvat entropy; and the Tsallis entropy.
is considered to be an additional element and it does not integrate the abstract idea into a practical application because the additional element is recited so generically (no details whatsoever are provided other than that it is a method wherein the one-parameter measure is selected from the following: the Renyi entropy; the Havrda-Charvat entropy; and the Tsallis entropy) that it represents no more than mere instructions to apply the judicial exception on a computer. As discussed in MPEP 2106.05(f), mere instructions to implement an abstract idea on a computer as a tool to perform an abstract idea is not indicative of integration into a practical application.
In the last step (Step 2B) of the analysis, the additional element does not amount to significantly more than the judicial exceptions. As explained with respect to Step 2A Prong Two, the method wherein the one-parameter measure is selected from the following: the Renyi entropy; the Havrda-Charvat entropy; and the Tsallis entropy, is at best the equivalent of merely adding the words “apply it” to the judicial exception. See MPEP 2106.05(f). Mere instructions to apply an exception cannot provide an inventive concept and does not amount to significantly more than the judicial exception. The claim is not patent eligible.
Regarding claim 23:
In the next step (Step 2A, prong 1) of the analysis, the limitation of:
wherein said selecting comprises selecting a plurality of machine-learning source domain candidates having respective diversity metrics above a predetermined threshold.
Under the broadest reasonable interpretation, the above limitations are process steps that cover mental processes including an observation, evaluation, judgment or opinion that could be performed in the mind or with the aid of pencil and paper but for the recitation of a generic computer component. If a claim, under its broadest reasonable interpretation, covers a mental process but for the recitation of generic computer components, then it falls within the “Mental Process” grouping of abstract ideas.
In the next step (Step 2A, prong 2) of the analysis, the limitation
and wherein said applying comprises applying each of the selected machine-learning source domain candidates to the target domain.
is considered to be an additional element and it does not integrate the abstract idea into a practical application because the additional element is recited so generically (no details whatsoever are provided other than that it is a method wherein said applying comprises applying each of the selected machine-learning source domain candidates to the target domain) that it represents no more than mere instructions to apply the judicial exception on a computer. As discussed in MPEP 2106.05(f), mere instructions to implement an abstract idea on a computer as a tool to perform an abstract idea is not indicative of integration into a practical application.
In the last step (Step 2B) of the analysis, the additional element does not amount to significantly more than the judicial exceptions. As explained with respect to Step 2A Prong Two, the method wherein said applying comprises applying each of the selected machine-learning source domain candidates to the target domain, is at best the equivalent of merely adding the words “apply it” to the judicial exception. See MPEP 2106.05(f). Mere instructions to apply an exception cannot provide an inventive concept and does not amount to significantly more than the judicial exception. The claim is not patent eligible.
Regarding claim 24:
In the next step (Step 2A, prong 1) of the analysis, the limitation of:
wherein said calculating, selecting, and applying comprises transfer learning performed in response to detecting a change in the execution environment of the target domain.
Under the broadest reasonable interpretation, the above limitations are process steps that cover mental processes including an observation, evaluation, judgment or opinion that could be performed in the mind or with the aid of pencil and paper but for the recitation of a generic computer component. If a claim, under its broadest reasonable interpretation, covers a mental process but for the recitation of generic computer components, then it falls within the “Mental Process” grouping of abstract ideas.
In the next step (Step 2A, prong 2) of the analysis, it does not integrate into a practical application because it does not add any additional elements that integrate the abstract idea into practical application.
In the last step (Step 2B) of the analysis, it does not add any additional elements that amount to significantly more than the abstract idea and thus fails to add an inventive concept. The claim is not patent eligible.
Regarding claim 25:
In the next step (Step 2A, prong 1) of the analysis, the limitation of:
wherein detecting the change in the execution environment comprises detecting a change in feature space in the target domain.
Under the broadest reasonable interpretation, the above limitations are process steps that cover mental processes including an observation, evaluation, judgment or opinion that could be performed in the mind or with the aid of pencil and paper but for the recitation of a generic computer component. If a claim, under its broadest reasonable interpretation, covers a mental process but for the recitation of generic computer components, then it falls within the “Mental Process” grouping of abstract ideas.
In the next step (Step 2A, prong 2) of the analysis, it does not integrate into a practical application because it does not add any additional elements that integrate the abstract idea into practical application.
In the last step (Step 2B) of the analysis, it does not add any additional elements that amount to significantly more than the abstract idea and thus fails to add an inventive concept. The claim is not patent eligible.
Regarding claim 26:
In the next step (Step 2A, prong 1) of the analysis, the limitation of:
wherein detecting the change in the execution environment comprises detecting a change in a machine-learning task in the target domain.
Under the broadest reasonable interpretation, the above limitations are process steps that cover mental processes including an observation, evaluation, judgment or opinion that could be performed in the mind or with the aid of pencil and paper but for the recitation of a generic computer component. If a claim, under its broadest reasonable interpretation, covers a mental process but for the recitation of generic computer components, then it falls within the “Mental Process” grouping of abstract ideas.
In the next step (Step 2A, prong 2) of the analysis, it does not integrate into a practical application because it does not add any additional elements that integrate the abstract idea into practical application.
In the last step (Step 2B) of the analysis, it does not add any additional elements that amount to significantly more than the abstract idea and thus fails to add an inventive concept. The claim is not patent eligible.
Regarding claim 27:
In the next step (Step 2A, prong 1) of the analysis, the limitation of:
wherein detecting the change in the execution environment comprises detecting a change in resources available in the execution environment.
Under the broadest reasonable interpretation, the above limitations are process steps that cover mental processes including an observation, evaluation, judgment or opinion that could be performed in the mind or with the aid of pencil and paper but for the recitation of a generic computer component. If a claim, under its broadest reasonable interpretation, covers a mental process but for the recitation of generic computer components, then it falls within the “Mental Process” grouping of abstract ideas.
In the next step (Step 2A, prong 2) of the analysis, it does not integrate into a practical application because it does not add any additional elements that integrate the abstract idea into practical application.
In the last step (Step 2B) of the analysis, it does not add any additional elements that amount to significantly more than the abstract idea and thus fails to add an inventive concept. The claim is not patent eligible.
Regarding claim 28:
In the next step (Step 2A, prong 2) of the analysis, the limitation
wherein said calculating, selecting, and applying is performed as part of inclusion in a federation for federated machine learning.
is considered to be an additional element and it does not integrate the abstract idea into a practical application because the additional element is recited so generically (no details whatsoever are provided other than that it is a method wherein said calculating, selecting, and applying is performed as part of inclusion in a federation for federated machine learning) that it represents no more than mere instructions to apply the judicial exception on a computer. As discussed in MPEP 2106.05(f), mere instructions to implement an abstract idea on a computer as a tool to perform an abstract idea is not indicative of integration into a practical application.
In the last step (Step 2B) of the analysis, the additional element does not amount to significantly more than the judicial exceptions. As explained with respect to Step 2A Prong Two, the method wherein said calculating, selecting, and applying is performed as part of inclusion in a federation for federated machine learning, is at best the equivalent of merely adding the words “apply it” to the judicial exception. See MPEP 2106.05(f). Mere instructions to apply an exception cannot provide an inventive concept and does not amount to significantly more than the judicial exception. The claim is not patent eligible.
Regarding claim 29:
In the next step (Step 2A, prong 1) of the analysis, the limitation of:
wherein identifying the plurality of machine-learning source domain candidates comprises comparing a feature space for each machine-learning source domain candidate to a feature space of the target domain.
Under the broadest reasonable interpretation, the above limitations are process steps that cover mental processes including an observation, evaluation, judgment or opinion that could be performed in the mind or with the aid of pencil and paper but for the recitation of a generic computer component. If a claim, under its broadest reasonable interpretation, covers a mental process but for the recitation of generic computer components, then it falls within the “Mental Process” grouping of abstract ideas.
In the next step (Step 2A, prong 2) of the analysis, it does not integrate into a practical application because it does not add any additional elements that integrate the abstract idea into practical application.
In the last step (Step 2B) of the analysis, it does not add any additional elements that amount to significantly more than the abstract idea and thus fails to add an inventive concept. The claim is not patent eligible.
Regarding claim 30:
In the next step (Step 2A, prong 2) of the analysis, the limitation
wherein the execution environment comprises one or more servers in a telecommunications network and applying the selected machine-learning source domain candidate comprises using the selected machine-learning source domain candidate for management of one or more telecommunications tasks in the telecommunications network.
is considered to be an additional element and it does not integrate the abstract idea into a practical application because the additional element is recited so generically (no details whatsoever are provided other than that it is a method wherein the execution environment comprises one or more servers in a telecommunications network and applying the selected machine-learning source domain candidate comprises using the selected machine-learning source domain candidate for management of one or more telecommunications tasks in the telecommunications network) that it represents no more than mere instructions to apply the judicial exception on a computer. As discussed in MPEP 2106.05(f), mere instructions to implement an abstract idea on a computer as a tool to perform an abstract idea is not indicative of integration into a practical application.
In the last step (Step 2B) of the analysis, the additional element does not amount to significantly more than the judicial exceptions. As explained with respect to Step 2A Prong Two, the method wherein the execution environment comprises one or more servers in a telecommunications network and applying the selected machine-learning source domain candidate comprises using the selected machine-learning source domain candidate for management of one or more telecommunications tasks in the telecommunications network, is at best the equivalent of merely adding the words “apply it” to the judicial exception. See MPEP 2106.05(f). Mere instructions to apply an exception cannot provide an inventive concept and does not amount to significantly more than the judicial exception. The claim is not patent eligible.
Regarding claims 31-37:
According to the first step (Step 1) of the 101 analysis, claims 31-37 are directed to a server node (manufacture) and falls within one of the four statutory categories (i.e., process, machine, manufacture, or composition of matter).
Regarding claim 31:
In step (Step 2A, prong 2) of the analysis, the limitation of:
A server node, comprising: communication circuitry configured for communication with one or more other nodes in a network; and processing circuitry configured to:
is considered to be an additional element and it does not integrate the abstract idea into a practical application because the additional element is recited so generically (no details whatsoever are provided other than that it is a server node, comprising: communication circuitry configured for communication with one or more other nodes in a network; and processing circuitry configured to perform something) that it represents no more than mere instructions to apply the judicial exception on a computer. As discussed in MPEP 2106.05(f), mere instructions to implement an abstract idea on a computer as a tool to perform an abstract idea is not indicative of integration into a practical application.
The rest of the limitations of claim 31 are substantially similar to claim 19 and therefore is rejected on similar grounds as claim 19 as explained above.
Regarding claim 32:
Claim 32 is substantially similar to claim 20 and therefore is rejected on similar grounds as claim 20.
Regarding claim 33:
Claim 33 is substantially similar to claim 21 and therefore is rejected on similar grounds as claim 21.
Regarding claim 34:
Claim 34 is substantially similar to claim 22 and therefore is rejected on similar grounds as claim 22.
Regarding claim 35:
Claim 35 is substantially similar to claim 23 and therefore is rejected on similar grounds as claim 23.
Regarding claim 36:
Claim 36 is substantially similar to claim 24 and therefore is rejected on similar grounds as claim 24.
Regarding claim 37:
Claim 37 is substantially similar to claims 25-27 and therefore is rejected on similar grounds as claim 25-27.
Regarding claim 38:
According to the first step (Step 1) of the 101 analysis, claim 38 is directed to a non-transitory computer-readable medium (manufacture) and falls within one of the four statutory categories (i.e., process, machine, manufacture, or composition of matter).
The rest of the limitations of claim 38 are substantially similar to claim 19 and therefore is rejected on similar grounds as claim 19 as explained above.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
Claims 19-21, 24-27, 29, 31-33, 36, 37, and 38 are rejected under 35 U.S.C. 103 as being unpatentable over Wu et al (Entropy Minimization vs. Diversity Maximization for Domain Adaptation, 2020) in view of Taylor et al (Transfer Learning for Reinforcement Learning Domains: A Survey, 2009).
Regarding claim 19:
Wu teaches: A method for machine-learning adaptation, the method comprising: identifying a plurality of machine-learning source domain candidates ([Page 1, Column 2, Last 2 Paragraphs] 1) We propose a minimal-entropy diversity maximization (MEDM) method for UDA. 2) MEDM outperforms state-of-the-art methods on four domain adaptation datasets, including VisDA-2017, ImageCLEF, Office-Home and Office-31. [Page 2] Figure 1. Note: Figure 1 shows a batch of source samples corresponding to source domain candidates and CNN (convolutional neural network) corresponds to Machine Learning);
calculating, for each of the identified machine-learning source domain candidates, a diversity metric, the diversity metric representing a marginalized measure of sample diversity of the respective machine-learning source domain candidate ([Page 3, Column 1, Section III] MINIMAL-ENTROPY DIVERSITY MAXIMIZATION A. Proposed Method: As the training of network is often implemented over batches of samples, the supervised loss for a given source batch S (for example, jSj = 32 for the batch size of 32) is accordingly modified as Ls(; S) =1jSjX(x;y)2S`(y; f(x)): [Page 3, Column 2, Paragraph 5] The use of EMO may produce trivial solutions as shown in Figure 1. By noting that a trivial solution shown in Figure 1 often has just one category, a nontrivial domain adaptation method may resort to producing sufficient category diversity in its solution. [Page 3, Column 2, Paragraph 6] In this paper, we employ the entropy of ˆq(T ) = [ˆq1, ˆq2, · · · , ˆqK] (6) for measuring the category diversity in a given target batch T . Formally, this category diversity over T can be measured as K Ld(θ,T ) H(ˆq(T )) = −k=1ˆqk log ˆqk. (8) [Page 3, Column 2, Paragraph 7] As this diversity metric does not require any priori information about the true category distribution q over Dt, its computation is easy to implement in practice. Note that random shuffling should be employed in training for maximizing (8). The objective of the proposed MEDM is to minES;T [Ls(; S) + Le(; T ) Ld(; T )] (9). [Page 7, Column 1, Paragraph 1] Office-Home [44] is a typical dataset with a large number of classes (65 classes), which containing 15,500 images from four visually very different domains: Artistic images, Clip Art, Product images, and Real-world images);
selecting the identified machine-learning source domain candidate having a highest diversity metric among the calculated diversity metrics ([Page 1, Column 2, Paragraph 5] In this paper, we make contributions towards close-to-perfect domain adaptation with entropy minimization. 1) We propose a minimal-entropy diversity maximization (MEDM) method for UDA. [Page 4, Column 1, Paragraph 9] With the use of diversity maximization, it may encourage to make prediction evenly across the batch, since the maximum value of Ld(θ∗,T ) could be achieved whenever q∗ = [1/K,··· ,1/K]. [Page 5, Column 1, Paragraph 2] When increases from 0, we would expect that the category diversity (8) increases correspondingly, which can help to avoid the trivial solutions. [Page 7, Column 1, Paragraph 3] although the category diversity is expected to achieve its maximum value when the inferred categories are uniformly-distributed. We guess that it works well due to the collaboration in meeting both requirements, namely, the minimization of entropy and the maximization of category diversity, where the parameter (9) is used to balance two individual requirements. [Conclusion] In this paper, we propose to employ diversity maximization for avoiding the trivial solutions. We show there exists a tradeoff for entropy minimization and diversity maximization towards the close-to-perfect domain adaptation. With the recently-proposed unsupervised model selection method, we show that the proposed MEDM outperforms state-of-the-art methods on several domain adaptation datasets, boosting a large margin especially on the largest VisDA dataset for cross-domain object classification. Note: Diversity maximization corresponds to highest diversity metric);
However, Wu does not explicitly disclose: and applying the selected machine-learning source domain candidate to a target domain in a new or changed execution environment.
Taylor teaches, in an analogous system: and applying the selected machine-learning source domain candidate to a target domain in a new or changed execution environment ([Page 1636, Paragraph 1] In this scenario, a total time scenario, which explicitly includes the time needed to learn the source task or tasks, would be most appropriate. On the other hand, a second reasonable goal of transfer is to effectively reuse past knowledge in a novel task. In this case, a target task time scenario, which only accounts for the time spent learning in the target task, is reasonable. Note: Novel task corresponds to a new or changed environment).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the system of Wu to incorporate the teachings of Taylor to apply the selected machine-learning source domain candidate to a target domain in a new or changed execution environment. One would have been motivated to do this modification because doing so would give the benefit of reducing the overall time required to learn a complex task as taught by Taylor [Page 1636, Paragraph 1].
Regarding claim 20:
The system of Wu and Taylor teaches: The method of claim 19 (as shown above).
Wu further teaches: wherein the diversity metric is calculated based on information theoretic measures ([Page 7, Column 2, Paragraph 3] We guess that it works well due to the collaboration in meeting both requirements, namely, the minimization of entropy and the maximization of category diversity, where the parameter B (9) is used to balance two individual requirements).
Regarding claim 21:
The system of Wu and Taylor teaches: The method of claim 20 (as shown above).
Wu further teaches: wherein the diversity metric is calculated based on a one- parameter measure of generalized entropy ([Page 7, Column 2, Paragraph 3] We guess that it works well due to the collaboration in meeting both requirements, namely, the minimization of entropy and the maximization of category diversity, where the parameter B (9) is used to balance two individual requirements. Note: Beta (B) corresponds to a one- parameter measure of generalized entropy).
Regarding claim 24:
The system of Wu and Taylor teaches: The method of claim 19 (as shown above).
Taylor further teaches: wherein said calculating, selecting, and applying comprises transfer learning performed in response to detecting a change in the execution environment of the target domain ([Page 1655, Last but one Paragraph] The idea of RTP is not only unique in this survey, but it is also potentially a very useful idea for transfer in general. While a number of TL methods are able to learn from a set of source tasks, no others attempt to automatically generate these source tasks. If the goal of an agent is perform as well as possible in a novel target task, it makes sense that the agent would try to train on many source tasks, even if they are artificial. How to best generate such source tasks so that they are most likely to be useful for an arbitrary target task in the same domain is an important area of open research. Note: Novel target task corresponds to detecting a change in the execution environment of the target domain).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the system of Wu to incorporate the teachings of Taylor wherein said calculating, selecting, and applying comprises transfer learning performed in response to detecting a change in the execution environment of the target domain. One would have been motivated to do this modification because doing so would give the benefit of reducing the overall time required to learn a complex task as taught by Taylor [Page 1636, Paragraph 1].
Regarding claim 25:
The system of Wu and Taylor teaches: The method of claim 24 (as shown above).
Taylor further teaches: wherein detecting the change in the execution environment comprises detecting a change in feature space in the target domain ([Page 1661, Last Paragraph] In our opinion, agent- and problem-space are ideas that should be further explored as they will likely yield additional benefits. Particularly in the case of physical agents, it is intuitive that agent sensors and actuators will be static, allowing information to be easily reused. Task-specific items [Page 1662, Paragraph 1] such as features and actions, may change, but should be faster to learn if the agent has already learned something about its unchanging agent-space).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the system of Wu to incorporate the teachings of Taylor wherein detecting the change in the execution environment comprises detecting a change in feature space in the target domain. One would have been motivated to do this modification because doing so would give the benefit of reducing the overall time required to learn a complex task as taught by Taylor [Page 1636, Paragraph 1].
Regarding claim 26:
The system of Wu and Taylor teaches: The method of claim 24 (as shown above).
Taylor further teaches: wherein detecting the change in the execution environment comprises detecting a change in a machine-learning task in the target domain ([Page 1639, Last but one Paragraph] For instance, if a target task is chosen that humans are relatively proficient at, transfer will provide them very little benefit. If that same target task is difficult for a machine learning algorithm, it will be relatively easy to show that the TL algorithm is quite effective relative to human transfer, even if the agent's absolute performance is extremely poor. [Page 1659, Last but one Paragraph] The learner's bias is important in all machine learning settings. However, Bayesian learning makes such bias explicit. Being able to set the bias through transfer from similar tasks may prove to be a very useful heuristic—we hope that additional transfer methods will be developed to initialize Bayesian learners from past tasks).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the system of Wu to incorporate the teachings of Taylor wherein detecting the change in the execution environment comprises detecting a change in a machine-learning task in the target domain. One would have been motivated to do this modification because doing so would give the benefit of reducing the overall time required to learn a complex task as taught by Taylor [Page 1636, Paragraph 1].
Regarding claim 27:
The system of Wu and Taylor teaches: The method of claim 24 (as shown above).
Taylor further teaches: wherein detecting the change in the execution environment comprises detecting a change in resources available in the execution environment ([Page 1641, Last Paragraph] The reward function, R : S → R, maps each state of the environment to a single number which is the instantaneous reward achieved for reaching the state. If the task is episodic, the agent begins at a start state and executes actions in the environment until it reaches a terminal state (one or more of the states in s final, which may be referred to as a goal state), at which point the agent is returned to a start state).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the system of Wu to incorporate the teachings of Taylor wherein detecting the change in the execution environment comprises detecting a change in resources available in the execution environment. One would have been motivated to do this modification because doing so would give the benefit of reducing the overall time required to learn a complex task as taught by Taylor [Page 1636, Paragraph 1].
Regarding claim 29:
The system of Wu and Taylor teaches: The method of claim 19 (as shown above).
Taylor further teaches: wherein identifying the plurality of machine-learning source domain candidates comprises comparing a feature space for each machine-learning source domain candidate to a feature space of the target domain ([Page 1662, Paragraph 2] For instance, in experiments the learner identified the concept of a fork, a state where the player could win on the subsequent turn regardless of what move the opponent took next. After training in the source task, analyzing the source task data for such features, and then setting the value for a given feature based on the source task data, such features of the game tree were used in a variety of target tasks. This analysis focuses on the effects of actions on the game tree and thus the actions and state variables describing the source and target game can differ without requiring an inter-task mapping).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the system of Wu to incorporate the teachings of Taylor wherein identifying the plurality of machine-learning source domain candidates comprises comparing a feature space for each machine-learning source domain candidate to a feature space of the target domain. One would have been motivated to do this modification because doing so would give the benefit of reducing the overall time required to learn a complex task as taught by Taylor [Page 1636, Paragraph 1].
Claims 23, 31-33, and 35-38 are rejected under 35 U.S.C. 103 as being unpatentable over Wu et al (Entropy Minimization vs. Diversity Maximization for Domain Adaptation, 2020) in view of Taylor et al (Transfer Learning for Reinforcement Learning Domains: A Survey, 2009) and further in view of Tan et al (US 20180240011 A1).
Regarding claim 23:
The system of Wu and Taylor teaches: The method of claim 19 (as shown above).
Wu further teaches: and wherein said applying comprises applying each of the selected machine-learning source domain candidates to the target domain ([Page 5, Column 2, Section B. VisDA-2017] The Visual Domain Adaption (VisDA) challenge [43] aims to test domain adaptation methods’s ability to transfer source knowledge and adapt it to novel target domains. As the largest domain-adaptation dataset, the VisDA dataset contains 280K images across 12 categories from the training, validation, and testing domains. The training domain (the source domain) is a set of synthetic 2D renderings of 3D models generated from different angles and with different lighting conditions, while the validation domain (the target domain) is a set of realistic photos. The source domain contains 152,397 synthetic images, and the target domain has 55,388 real images).
However, the system of Wu and Taylor does not explicitly disclose: wherein said selecting comprises selecting a plurality of machine-learning source domain candidates having respective diversity metrics above a predetermined threshold.
Tan teaches, in an analogous system: wherein said selecting comprises selecting a plurality of machine-learning source domain candidates having respective diversity metrics above a predetermined threshold ([0032] As noted, FIG. 2A illustrates a specific example in which the badge reader 212 provides the label 223. It is to be appreciated that labels may be generated in other manners and that the label does not need to be authoritative, or from a different data source. [0038] The sampler 224 forwards correctly predicted data-label pairs where the output entropy value H exceeds a predetermined threshold. Other than entropy, the forwarding decision can also be made based on alternative functions of the probability distribution, such as diversity index).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the combined system of Wu and Taylor to incorporate the teachings of Tan wherein said selecting comprises selecting a plurality of machine-learning source domain candidates having respective diversity metrics above a predetermined threshold. One would have been motivated to do this modification because doing so would give the benefit of forwarding correctly predicted data-label pairs as taught by Tan [0038].
Regarding claim 31:
Tan teaches: A server node, comprising: communication circuitry configured for communication with one or more other nodes in a network; and processing circuitry configured to ([0084] FIG. 13 is a block diagram of a computing device 1380 configured to implement the model merging techniques presented herein. That is, FIG. 13 illustrates one arrangement for a central learning unit (e.g., 116, 616) in accordance with examples presented herein. The computing device 1380 includes a network interface unit 1381 to enable network communications, one or more processors 1382, and memory 1383. The memory 1383 stores software modules that include local model instantiation logic 1350, a central machine learning model 1318, and a data and label generator 1325. These software modules, when executed by the one or more processors 577, causes the one or more processors to perform the operations described herein with reference to a central machine learning unit. Note: Central machine learning model corresponds to server node and processor corresponds to processing circuitry).
The rest of the limitations of claim 31 are substantially similar to claim 19 and therefore are rejected on similar grounds as claim 19 as explained above.
Regarding claim 32:
Claim 32 is substantially similar to claim 20 and therefore is rejected on similar grounds as claim 20.
Regarding claim 33:
Claim 33 is substantially similar to claim 21 and therefore is rejected on similar grounds as claim 21.
Regarding claim 35:
Claim 35 is substantially similar to claim 23 and therefore is rejected on similar grounds as claim 23.
Regarding claim 36:
Claim 36 is substantially similar to claim 24 and therefore is rejected on similar grounds as claim 24.
Regarding claim 37:
Claim 37 is substantially similar to claims 25-27 and therefore is rejected on similar grounds as claim 25-27.
Regarding claim 38:
Tan teaches: A non-transitory computer-readable medium comprising, stored thereupon, a computer program comprising instructions configured to cause a server executing the instructions to: ([0055] Thus, in general, the memory 578 may comprise one or more tangible (non-transitory) computer readable storage media (e.g., a memory device) encoded with software comprising computer executable instructions and when the software is executed (by the controller) it is operable to perform the operations described herein).
The rest of the limitations of claim 38 are substantially similar to claim 19 and therefore are rejected on similar grounds as claim 19 as explained above.
Claim 22 is rejected under 35 U.S.C. 103 as being unpatentable over Wu et al (Entropy Minimization vs. Diversity Maximization for Domain Adaptation, 2020) in view of Taylor et al (Transfer Learning for Reinforcement Learning Domains: A Survey, 2009) and further in view of Zhang et al (Preclinical Diagnosis of Magnetic Resonance (MR) Brain Images via Discrete Wavelet Packet Transform with Tsallis Entropy and Generalized Eigenvalue Proximal Support Vector Machine (GEPSVM), 2015).
Regarding claim 22:
The system of Wu and Taylor teaches: The method of claim 21 (as shown above).
However, the system of Wu and Taylor does not explicitly disclose: wherein the one-parameter measure is selected from the following: the Renyi entropy; the Havrda-Charvat entropy; and the Tsallis entropy.
Zhang teaches, in an analogous system: wherein the one-parameter measure is selected from the following: the Renyi entropy; the Havrda-Charvat entropy; and the Tsallis entropy ([Abstract] Tsallis entropy (TE) were harnessed).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the combined system of Wu and Taylor to incorporate the teachings of Zhang wherein the one-parameter measure is selected from the following: the Renyi entropy; the Havrda-Charvat entropy; and the Tsallis entropy. One would have been motivated to do this modification because doing so would give the benefit of harnessing Tsallis entropy to obtain entropy features as taught by Zhang [Abstract].
Claim 34 is rejected under 35 U.S.C. 103 as being unpatentable over Wu et al (Entropy Minimization vs. Diversity Maximization for Domain Adaptation, 2020) in view of Taylor et al (Transfer Learning for Reinforcement Learning Domains: A Survey, 2009) and Tan et al (US 20180240011 A1) and further in view of Zhang et al (Preclinical Diagnosis of Magnetic Resonance (MR) Brain Images via Discrete Wavelet Packet Transform with Tsallis Entropy and Generalized Eigenvalue Proximal Support Vector Machine (GEPSVM), 2015).
Regarding claim 34:
Claim 34 is substantially similar to claim 22 and therefore is rejected on similar grounds as claim 22.
Claim 28 is rejected under 35 U.S.C. 103 as being unpatentable over Wu et al (Entropy Minimization vs. Diversity Maximization for Domain Adaptation, 2020) in view of Taylor et al (Transfer Learning for Reinforcement Learning Domains: A Survey, 2009) and further in view of Milton (US 20200017117 A1).
Regarding claim 28:
The system of Wu and Taylor teaches: The method of claim 24 (as shown above).
However, the system of Wu and Taylor does not explicitly disclose: wherein said calculating, selecting, and applying is performed as part of inclusion in a federation for federated machine learning.
Milton teaches, in an analogous system: wherein said calculating, selecting, and applying is performed as part of inclusion in a federation for federated machine learning ([0099] Some embodiments may implement transfer learning using a federated framework, which can be labeled as a federated transfer learning (FTL) method).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the combined system of Wu and Taylor to incorporate the teachings of Milton wherein said calculating, selecting, and applying is performed as part of inclusion in a federation for federated machine learning. One would have been motivated to do this modification because doing so would give the benefit of an architecture that supports active learning to infer things about vehicles, drivers, and places based on relatively high-bandwidth on-board and road-side sensor feeds as taught by Milton [0017].
Claim 30 is rejected under 35 U.S.C. 103 as being unpatentable over Wu et al (Entropy Minimization vs. Diversity Maximization for Domain Adaptation, 2020) in view of Taylor et al (Transfer Learning for Reinforcement Learning Domains: A Survey, 2009) and further in view of Deasy et al (WO 2020028382 A1).
Regarding claim 30:
The system of Wu and Taylor teaches: The method of claim 19 (as shown above).
However, the system of Wu and Taylor does not explicitly disclose: wherein the execution environment comprises one or more servers in a telecommunications network and applying the selected machine-learning source domain candidate comprises using the selected machine-learning source domain candidate for management of one or more telecommunications tasks in the telecommunications network.
Deasy teaches, in an analogous system: wherein the execution environment comprises one or more servers in a telecommunications network and applying the selected machine-learning source domain candidate comprises using the selected machine-learning source domain candidate for management of one or more telecommunications tasks in the telecommunications network ([0099] Some embodiments may implement transfer learning using a federated framework, which can be labeled as a federated transfer learning (FTL) method).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the combined system of Wu and Taylor to incorporate the teachings of Deasy wherein the execution environment comprises one or more servers in a telecommunications network and applying the selected machine-learning source domain candidate comprises using the selected machine-learning source domain candidate for management of one or more telecommunications tasks in the telecommunications network. One would have been motivated to do this modification because doing so would give the benefit of allowing more efficient use of server resources as taught by Deasy [00299].
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Vogt et al (Unsupervised Source Selection for Domain Adaptation, 2018) discloses a fast domain similarity measure that captures the relatedness between datasets purely based on unlabeled data. Our method transfers knowledge from multiple sources by generating a weighted combination of domains. We show for multiple datasets that learning on such sources achieves an average overall accuracy closer than 2.5 percent to the results of the target classifier for semantic segmentation tasks. We further apply our method to the task of choosing informative patches from unlabeled datasets. Only labeling these patches enables a reduction in manual work of up to 85 percent.
Bascol et al (Improving Domain Adaptation By Source Selection, 2019) discloses domain adaptation for image classification with deep learning in the context of multiple available source domains. We propose a multisource domain adaptation method that selects and weights the sources based on inter-domain distances. We provide encouraging results on both classical benchmarks and a new real world application with 21 domains.
THIS ACTION IS MADE FINAL. Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to CHAITANYA RAMESH JAYAKUMAR whose telephone number is (571)272-3369. The examiner can normally be reached Mon-Fri 9am-1pm.
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, Omar Fernandez Rivas can be reached at (571)272-2589. 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.
/C.R.J./Examiner, Art Unit 2128
/OMAR F FERNANDEZ RIVAS/Supervisory Patent Examiner, Art Unit 2128
Prosecution Insights