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 Arguments
Remarks page 9-10, Applicant contends:
Claims 1 and 12 are not directed to abstract ideas
Response:
Claims 1 and 12 are considered directed to abstract ideas. Limitations, such as those noted in previous 101 claim rejections in previous office action, are indicated to perform elements completable in the human mind or with pen and paper (MPEP 2106.04(Section 3)). The claims are not seen as being integrated into a practical application via an improvement to a computing device, as the supposed improvement to a computing device is not apparent from the claims (MPEP 2106.05(a)). The claim limitations for claim 1 do not indicate any particular form of training, just where the training takes place based on a condition. Without further limitations on how the training taking place in one place (such as offline) versus another (such as online) is improving a computing device is not known. The paragraphs indicated to support the improvement in the specification (paragraphs 133-134) don’t appear to give an indication on how a computer or machine learning model is improved. The paragraphs indicate some methods can be slow in some cases, but how the claimed limitations provide a speedier system is not apparent.
MPEP sections referenced above:
MPEP 2106.04(Section 3): "The courts consider a mental process (thinking) that "can be performed in the human mind, or by a human using a pen and paper" to be an abstract idea. CyberSource Corp. v. Retail Decisions, Inc., 654 F.3d 1366, 1372, 99 USPQ2d 1690, 1695 (Fed. Cir. 2011). As the Federal Circuit explained, "methods which can be performed mentally, or which are the equivalent of human mental work, are unpatentable abstract ideas the ‘basic tools of scientific and technological work’ that are open to all.’" 654 F.3d at 1371, 99 USPQ2d at 1694 (citing Gottschalk v. Benson, 409 U.S. 63, 175 USPQ 673 (1972)). See also Mayo Collaborative Servs. v. Prometheus Labs. Inc., 566 U.S. 66, 71, 101 USPQ2d 1961, 1965 (2012) ("‘[M]ental processes[] and abstract intellectual concepts are not patentable, as they are the basic tools of scientific and technological work’" (quoting Benson, 409 U.S. at 67, 175 USPQ at 675)); Parker v. Flook, 437 U.S. 584, 589, 198 USPQ 193, 197 (1978) (same)."
MPEP 2106.05(a): "After the examiner has consulted the specification and determined that the disclosed invention improves technology, the claim must be evaluated to ensure the claim itself reflects the disclosed improvement in technology. Intellectual Ventures I LLC v. Symantec Corp., 838 F.3d 1307, 1316, 120 USPQ2d 1353, 1359 (Fed. Cir. 2016) (patent owner argued that the claimed email filtering system improved technology by shrinking the protection gap and mooting the volume problem, but the court disagreed because the claims themselves did not have any limitations that addressed these issues). That is, the claim must include the components or steps of the invention that provide the improvement described in the specification. However, the claim itself does not need to explicitly recite the improvement described in the specification (e.g., "thereby increasing the bandwidth of the channel"). The full scope of the claim under the BRI should be considered to determine if the claim reflects an improvement in technology (e.g., the improvement described in the specification). In making this determination, it is critical that examiners look at the claim "as a whole," in other words, the claim should be evaluated "as an ordered combination, without ignoring the requirements of the individual steps." When performing this evaluation, examiners should be "careful to avoid oversimplifying the claims" by looking at them generally and failing to account for the specific requirements of the claims. McRO, 837 F.3d at 1313, 120 USPQ2d at 1100."
Remarks page 11, applicant contends:
Claims are directed to a practical application of an abstract idea as a result of being amended to overcome 102 and 103 rejections.
Response:
101 rejections are judged separate to 103 rejections. While not being obvious under prior art can be an indication of possibly being not well understood, routine, or conventional, the absence of 103 rejections does not mean the claim limitations satisfy the requirements under 101 (MPEP 2106.05(1)).
MPEP 2106.05(1): "Although the courts often evaluate considerations such as the conventionality of an additional element in the eligibility analysis, the search for an inventive concept should not be confused with a novelty or non-obviousness determination. See Mayo, 566 U.S. at 91, 101 USPQ2d at 1973 (rejecting "the Government’s invitation to substitute §§ 102, 103, and 112 inquiries for the better established inquiry under § 101 "). As made clear by the courts, the "‘novelty’ of any element or steps in a process, or even of the process itself, is of no relevance in determining whether the subject matter of a claim falls within the § 101 categories of possibly patentable subject matter." Intellectual Ventures I v. Symantec Corp., 838 F.3d 1307, 1315, 120 USPQ2d 1353, 1358 (Fed. Cir. 2016) (quoting Diamond v. Diehr, 450 U.S. at 188–89, 209 USPQ at 9)…"
As noted in above responses to remarks, the supposed improvement to computing is not apparent from the claims, thus is not seen as indicating a practical application. The current claim limitations are not seen as invoking a similar situation to Desjardins, as Desjardins notes a way of updating parameters of a machine learning model to help preserve functionality while training to improve other functionality. No indication is given in claim 1 (or analogous claim 12) of a particular method for improving a computing system or machine learning model that is akin to the situation in Desjardins.
The applicant’s arguments are not seen as convincing, thus the 101 rejections are maintained.
Remarks page 12-15, applicant contends:
The combination of Tanimoto and Gottschlitch fail to discloses all of the limitations set forth in the independent claims.
Response:
The limitations set for in the independent claims are seen as taught by Tanimoto and Gottschlitch (at least in regards to claim 1). Tanimoto teaches aspects related to calculating the difference between distributions, inference dataset, second dataset, and other aspects noted in the previous claim 1 rejection. Gottschlitch, as noted by previous claim 4 rejection, taught the utilization of a condition for determining whether to train online or offline based on a difference. To provide further clarification on how Gottschlitch is considered teaching or being able to be used in combination with the difference between sets, figure 12 of the current application is deemed the support for what the difference between datasets is considered by the current application and figure 8 of Gottschlitch shows that the steps performed fit the description. Figure 12 of the current application shows predictions are compared to determine a prediction error which is then output as the data distribution difference. Figure 8 810 of Gottschlitch shows that convergence is determined by a comparison between the predictions (which can be considered giving a prediction error) which is then used to determine whether to perform an offline or online update (further supported by the quotes utilized in previous claim 4). Thus Gottschlitch is considered calculating a difference between data distributions. No particular method is described in the claims for how the difference between the data distributions should be calculated. The claim language is also broad in terms of what the offline and online condition are, which under BRI means elements potentially not intended to be referred to by the applicant fit the claim limitations. Aspects related to the details of a dataset being an inference dataset or training dataset are originally taught by Tanimoto in previous claim 1.
A possible method to overcome the issue is to add details pertaining to the noted improvements by applicant within the remarks (which can also help with 101 interpretation). Page 11 of remarks notes “For instance, claims 1-3, 5-14, and 16-20 improve the performance of a computing device in training an AI model by avoiding slow online updates when the distribution shift is large and avoiding unnecessary time-consuming offline retraining when the distribution shift is small”. Such an improvement is not apparent within the claim limitations, but if details within were added to the claims (that are supported by the specification) that support the difference and condition are setup to avoid things like “unnecessary time-consuming updates” could likely overcome aspects of Gottschlitch, as Gottschlitch is unlikely to contain more specific details of the current application. Other possible ideas could be providing more descriptive aspects of what the difference must contain, as that could present an amendment that results in the difference found in Gottschlitch no longer fitting the difference wanted in the claim limitations under BRI.
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 1-3, 5-14, 16-22 are rejected under 35 U.S.C. 101 because the claimed invention is directed towards an abstract idea without significantly more.
In regards to Claim 1:
Step 1: Is the claim directed towards a process, machine, manufacture, or composition of matter?
Yes, the claim is directed towards a method, so a process.
Step 2A Prong 1: Does the claim recite a law of nature, a natural phenomenon, or an abstract idea?
Yes, the claim does recite a(n) abstract idea.
Claim 1 recites the following abstract ideas:
in response to a difference between a first data distribution of the inference data set and a second data distribution of a training data set meeting an offline update condition
This limitation is directed towards the abstract idea of a mental process, or a concept performed in the human mind, including observation, evaluation, judgement or opinion (see MPEP 2106.04(a)(2) subsection 3). Here the limitation is seen as evaluation and judgement.
in response to a difference between the first data distribution of the inference data set and the second data distribution of the training data set not meeting the offline update condition
This limitation is directed towards the abstract idea of a mental process, or a concept performed in the human mind, including observation, evaluation, judgement or opinion (see MPEP 2106.04(a)(2) subsection 3). Here the limitation is seen as evaluation and judgement.
Step 2A Prong 2: Does the claim recite additional elements that integrate the exception into a practical application of the exception?
No, the application does not recite any additional elements that would integrate the abstract idea into a practical application.
Claim 1 recites the following additional elements:
obtaining an inference data set comprising inference data, wherein the inference data is for inputting into an existing artificial intelligence (AI) model for performing inference;
This limitation is directed towards the insignificant extra solution activity of mere data gathering (see MPEP § 2106.05(g)).
training, by using the inference data set…, the existing Al model offline to obtain an updated AI model
At a high level of generality, this is an activity of using the inference data set as an “apply it” use (see MPEP 2106.05(f)).
training, by using the inference data set…, the existing Al model online to obtain an updated AI model
At a high level of generality, this is an activity of using the inference data set as an “apply it” use (see MPEP 2106.05(f)).
Step 2B: Does the claim as a whole amount to significantly more than the judicial exception?
No, the claim as a whole does not amount to significantly more than the judicial exception. All elements of the claim, viewed individually or wholistically, do not provide an inventive concept or otherwise significantly more than the abstract idea itself.
Claim 1 recites the following additional elements:
obtaining an inference data set comprising inference data, wherein the inference data is for inputting into an existing artificial intelligence (AI) model for performing inference;
This limitation is directed towards the insignificant extra solution activity of mere data gathering (see MPEP § 2106.05(g)). This is a well understood, routine, conventional activity of transmitting data (see MPEP 2106.05(d) example i in computer functions).
training, by using the inference data set…, the existing Al model offline to obtain an updated AI model
At a high level of generality, this is an activity of using the inference data set as an “apply it” use (see MPEP 2106.05(f)). At said high level of generality, a generic recitation of “training, using the inference data set…” the existing AI model does not incorporate the abstract idea into a practical invention and is seen as a variation of the phrase “apply it”.
training, by using the inference data set…, the existing Al model online to obtain an updated AI model
At a high level of generality, this is an activity of using the inference data set as an “apply it” use (see MPEP 2106.05(f)). At said high level of generality, a generic recitation of “training, using the inference data set…” the existing AI model does not incorporate the abstract idea into a practical invention and is seen as a variation of the phrase “apply it”.
In regards to Claim 2:
Step 2A Prong 1: Does the claim recite a law of nature, a natural phenomenon, or an abstract idea?
Yes, the claim does recite a(n) abstract idea.
Claim 2 recites the following abstract ideas:
wherein the existing AI model is deployed on an inference platform,
This limitation is directed towards the continuation of the abstract ideas of a mental process, or a concept performed in the human mind, including observation, evaluation, judgement or opinion (see MPEP 2106.04(a)(2) subsection 3) from claim 1.
and wherein the method further comprises: comparing a first inference precision of the updated Al model with a second inference precision of the existing AI model
This limitation is directed towards the abstract idea of a mental process, or a concept performed in the human mind, including observation, evaluation, judgement or opinion (see MPEP 2106.04(a)(2) subsection 3). Here the limitation is seen as evaluation. This limitation is seen as comparing two numbers.
Step 2A Prong 2: Does the claim recite additional elements that integrate the exception into a practical application of the exception?
No, the application does not recite any additional elements that would integrate the abstract idea into a practical application.
Claim 2 recites the following additional elements:
and deploying, in response to the first inference precision being higher than the second inference precision, the updated Al model on the inference platform to perform inference in place of the existing Al model
This limitation is directed towards the insignificant extra solution activity of mere data gathering (see MPEP § 2106.05(g)).
Step 2B: Does the claim as a whole amount to significantly more than the judicial exception?
No, the claim as a whole does not amount to significantly more than the judicial exception. All elements of the claim, viewed individually or wholistically, do not provide an inventive concept or otherwise significantly more than the abstract idea itself.
Claim 2 recites the following additional elements:
and deploying, in response to the first inference precision being higher than the second inference precision, the updated Al model on the inference platform to perform inference in place of the existing Al model
This limitation is directed towards the insignificant extra solution activity of mere data gathering (see MPEP § 2106.05(g)). This is a well understood, routine, conventional activity of transmitting data (see MPEP 2106.05(d) example i in computer functions).
In regards to Claim 3:
Step 2A Prong 2: Does the claim recite additional elements that integrate the exception into a practical application of the exception?
No, the application does not recite any additional elements that would integrate the abstract idea into a practical application.
Claim 3 recites the following additional elements:
wherein before deploying the updated Al model, the method further comprises: displaying, by using a display interface, the first inference precision and the second inference precision;
This limitation is directed towards the insignificant extra solution activity of mere data gathering (see MPEP § 2106.05(g)).
and receiving, from a user in response to displaying the first inference precision and the second inference precision, an update instruction for the existing Al model
This limitation is directed towards the insignificant extra solution activity of mere data gathering (see MPEP § 2106.05(g)).
Step 2B: Does the claim as a whole amount to significantly more than the judicial exception?
No, the claim as a whole does not amount to significantly more than the judicial exception. All elements of the claim, viewed individually or wholistically, do not provide an inventive concept or otherwise significantly more than the abstract idea itself.
Claim 3 recites the following additional elements:
wherein before deploying the updated Al model, the method further comprises: displaying, by using a display interface, the first inference precision and the second inference precision;
This limitation is directed towards the insignificant extra solution activity of mere data gathering (see MPEP § 2106.05(g)). This is a well understood, routine, conventional activity of transmitting data (see MPEP 2106.05(d) example i in computer functions).
and receiving, from a user in response to displaying the first inference precision and the second inference precision, an update instruction for the existing Al model
This limitation is directed towards the insignificant extra solution activity of mere data gathering (see MPEP § 2106.05(g)). This is a well understood, routine, conventional activity of transmitting data (see MPEP 2106.05(d) example i in computer functions).
In regards to Claim 5:
Step 2A Prong 1: Does the claim recite a law of nature, a natural phenomenon, or an abstract idea?
Yes, the claim does recite a(n) abstract idea.
Claim 5 recites the following abstract ideas:
calculating, by using the difference, a parameter change amount of a first target part in the existing AI model
This limitation is directed towards the abstract idea of a mental process, or a concept performed in the human mind, including observation, evaluation, judgement or opinion (see MPEP 2106.04(a)(2) subsection 3). Here the limitation is seen as evaluation. Without more detail on what a parameter change requires and without any adjustment actually occurring to parameters, this limitation is interpreted as generating a number (a parameter change amount) based on another number (difference).
and calculating, based on a current parameter and the parameter change amount, a parameter of a second target part in the updated AI model
This limitation is directed towards the abstract idea of a mental process, or a concept performed in the human mind, including observation, evaluation, judgement or opinion (see MPEP 2106.04(a)(2) subsection 3). Here the limitation is seen as evaluation. Without more details on how the determining is done and what exactly a parameter is this limitation is seen as determining/generating a number (a parameter of a second part) based on some other numbers (a current parameter and the parameter change amount).
In regards to Claim 6:
Step 2A Prong 2: Does the claim recite additional elements that integrate the exception into a practical application of the exception?
No, the application does not recite any additional elements that would integrate the abstract idea into a practical application.
Claim 6 recites the following additional elements:
constructing, based on the inference data set, a target data set;
At a high level of generality, this is an activity of using the inference data set as an “apply it” use (see MPEP 2106.05(f)). At said high level of generality, a generic recitation of “constructing… a target data set” based on the inference data set does not incorporate the abstract idea into a practical invention and is seen as a variation of the phrase “apply it”.
and updating, by using the target data set, the existing Al model
At a high level of generality, this is an activity of using the inference data set as an “apply it” use (see MPEP 2106.05(f)). At said high level of generality, a generic recitation of “updating, using the inference data set…” the existing AI model does not incorporate the abstract idea into a practical invention and is seen as a variation of the phrase “apply it”.
Step 2B: Does the claim as a whole amount to significantly more than the judicial exception?
No, the claim as a whole does not amount to significantly more than the judicial exception. All elements of the claim, viewed individually or wholistically, do not provide an inventive concept or otherwise significantly more than the abstract idea itself.
Claim 6 recites the following additional elements:
constructing, based on the inference data set, a target data set;
At a high level of generality, this is an activity of using the inference data set as an “apply it” use (see MPEP 2106.05(f)). At said high level of generality, a generic recitation of “constructing… a target data set” based on the inference data set does not incorporate the abstract idea into a practical invention and is seen as a variation of the phrase “apply it”.
and updating, by using the target data set, the existing Al model
At a high level of generality, this is an activity of using the inference data set as an “apply it” use (see MPEP 2106.05(f)). At said high level of generality, a generic recitation of “updating, using the inference data set…” the existing AI model does not incorporate the abstract idea into a practical invention and is seen as a variation of the phrase “apply it”.
In regards to Claim 7:
Step 2A Prong 2: Does the claim recite additional elements that integrate the exception into a practical application of the exception?
No, the application does not recite any additional elements that would integrate the abstract idea into a practical application.
Claim 7 recites the following additional elements:
obtaining, from the inference data set, target data that meets a sample condition;
This limitation is directed towards the insignificant extra solution activity of mere data gathering (see MPEP § 2106.05(g)).
displaying, by using a display interface, the target data;
This limitation is directed towards the insignificant extra solution activity of mere data gathering (see MPEP § 2106.05(g)).
obtaining, from a user, a result of labeling the target data;
This limitation is directed towards the insignificant extra solution activity of mere data gathering (see MPEP § 2106.05(g)).
and constructing, based on the target data and the result, the target data set
At a high level of generality, this is an activity of using the target data and the result as an “apply it” use (see MPEP 2106.05(f)).
Step 2B: Does the claim as a whole amount to significantly more than the judicial exception?
No, the claim as a whole does not amount to significantly more than the judicial exception. All elements of the claim, viewed individually or wholistically, do not provide an inventive concept or otherwise significantly more than the abstract idea itself.
Claim 7 recites the following additional elements:
obtaining, from the inference data set, target data that meets a sample condition;
This limitation is directed towards the insignificant extra solution activity of mere data gathering (see MPEP § 2106.05(g)). This is a well understood, routine, conventional activity of transmitting data (see MPEP 2106.05(d) example i in computer functions).
displaying, by using a display interface, the target data;
This limitation is directed towards the insignificant extra solution activity of mere data gathering (see MPEP § 2106.05(g)). This is a well understood, routine, conventional activity of transmitting data (see MPEP 2106.05(d) example i in computer functions).
obtaining, from a user, a result of labeling the target data;
This limitation is directed towards the insignificant extra solution activity of mere data gathering (see MPEP § 2106.05(g)). This is a well understood, routine, conventional activity of transmitting data (see MPEP 2106.05(d) example i in computer functions).
and constructing, based on the target data and the result, the target data set
At a high level of generality, this is an activity of using the target data and the result as an “apply it” use (see MPEP 2106.05(f)). At said high level of generality, a generic recitation of “constructing… the target data set” based on the target data and the result does not incorporate the abstract idea into a practical invention and is seen as a variation of the phrase “apply it”.
In regards to Claim 8:
Step 2A Prong 2: Does the claim recite additional elements that integrate the exception into a practical application of the exception?
No, the application does not recite any additional elements that would integrate the abstract idea into a practical application.
Claim 8 recites the following additional elements:
wherein the target data set comprises unlabeled data and labeled data,
At a high level of generality, this is a continuation of an activity of using the inference data set as an “apply it” use (see MPEP 2106.05(f)).
and wherein updating the existing Al model comprises: optimizing, by using the unlabeled data, a feature extraction part in the existing Al model in an unsupervised manner to produce an optimized feature extraction part;
At a high level of generality, this is an activity of using unlabeled data as an “apply it” use (see MPEP 2106.05(f)).
and updating, based on the optimized feature extraction part and the labeled data, the existing Al model
At a high level of generality, this is an activity of using “the optimized feature extraction part and the labeled data” as an “apply it” use (see MPEP 2106.05(f)).
Step 2B: Does the claim as a whole amount to significantly more than the judicial exception?
No, the claim as a whole does not amount to significantly more than the judicial exception. All elements of the claim, viewed individually or wholistically, do not provide an inventive concept or otherwise significantly more than the abstract idea itself.
Claim 8 recites the following additional elements:
wherein the target data set comprises unlabeled data and labeled data,
At a high level of generality, this is a continuation of an activity of using the inference data set as an “apply it” use (see MPEP 2106.05(f)). At said high level of generality, a recitation of the target dataset comprising unlabeled data and labeled data does not incorporate the abstract idea into a practical invention and is seen as a continuation of the “apply it” in claim 6.
and wherein updating the existing Al model comprises: optimizing, by using the unlabeled data, a feature extraction part in the existing Al model in an unsupervised manner to produce an optimized feature extraction part;
At a high level of generality, this is an activity of using unlabeled data as an “apply it” use (see MPEP 2106.05(f)). At said high level of generality, a generic recitation of “optimizing… a feature extraction part in the existing AI model” does not incorporate the abstract idea into a practical invention and is seen as a variation of the phrase “apply it”. Optimizing is seen as a form of training in this limitation.
and updating, based on the optimized feature extraction part and the labeled data, the existing Al model
At a high level of generality, this is an activity of using “the optimized feature extraction part and the labeled data” as an “apply it” use (see MPEP 2106.05(f)). At said high level of generality, a generic recitation of “updating, based on the optimized feature extraction part and the labeled data” the existing AI model does not incorporate the abstract idea into a practical invention and is seen as a variation of the phrase “apply it”.
In regards to Claim 9:
Step 2A Prong 2: Does the claim recite additional elements that integrate the exception into a practical application of the exception?
No, the application does not recite any additional elements that would integrate the abstract idea into a practical application.
Claim 9 recites the following additional elements:
target data set comprises unlabeled data and labeled data,
At a high level of generality, this is a continuation of an activity of using the inference data set as an “apply it” use (see MPEP 2106.05(f)).
and wherein updating the existing Al model comprises: labeling, by using the existing Al model, the unlabeled data to obtain a labeling result of the unlabeled data;
At a high level of generality, this is an activity of using “the existing AI model” as an “apply it” use (see MPEP 2106.05(f)).
and updating, based on the labeling result and the labeled data, the existing Al model
At a high level of generality, this is an activity of using “the labeling result and the labeled data” as an “apply it” use (see MPEP 2106.05(f)).
Step 2B: Does the claim as a whole amount to significantly more than the judicial exception?
No, the claim as a whole does not amount to significantly more than the judicial exception. All elements of the claim, viewed individually or wholistically, do not provide an inventive concept or otherwise significantly more than the abstract idea itself.
Claim 9 recites the following additional elements:
target data set comprises unlabeled data and labeled data,
At a high level of generality, this is a continuation of an activity of using the inference data set as an “apply it” use (see MPEP 2106.05(f)). At said high level of generality, a recitation of the target dataset comprising unlabeled data and labeled data does not incorporate the abstract idea into a practical invention and is seen as a continuation of the “apply it” in claim 6.
and wherein updating the existing Al model comprises: labeling, by using the existing Al model, the unlabeled data to obtain a labeling result of the unlabeled data;
At a high level of generality, this is an activity of using “the existing AI model” as an “apply it” use (see MPEP 2106.05(f)). At said high level of generality, a generic recitation of “labeling… the unlabeled data” using the existing AI model does not incorporate the abstract idea into a practical invention and is seen as a variation of the phrase “apply it”.
and updating, based on the labeling result and the labeled data, the existing Al model
At a high level of generality, this is an activity of using “the labeling result and the labeled data” as an “apply it” use (see MPEP 2106.05(f)). At said high level of generality, a generic recitation of “updating, based on the labeling result and the labeled data” the existing AI model does not incorporate the abstract idea into a practical invention and is seen as a variation of the phrase “apply it”.
In regards to Claim 10:
Step 2A Prong 2: Does the claim recite additional elements that integrate the exception into a practical application of the exception?
No, the application does not recite any additional elements that would integrate the abstract idea into a practical application.
Claim 10 recites the following additional elements:
obtaining, based on a data characteristic of data in the target data set, a policy for updating the existing Al model to obtain an updated policy
This limitation is directed towards the insignificant extra solution activity of mere data gathering (see MPEP § 2106.05(g)).
and updating, based on the updated policy, the existing Al model
At a high level of generality, this is an activity of using “the updated policy” as an “apply it” use (see MPEP 2106.05(f)).
Step 2B: Does the claim as a whole amount to significantly more than the judicial exception?
No, the claim as a whole does not amount to significantly more than the judicial exception. All elements of the claim, viewed individually or wholistically, do not provide an inventive concept or otherwise significantly more than the abstract idea itself.
Claim 10 recites the following additional elements:
obtaining, based on a data characteristic of data in the target data set, a policy for updating the existing Al model to obtain an updated policy
This limitation is directed towards the insignificant extra solution activity of mere data gathering (see MPEP § 2106.05(g)). This is a well understood, routine, conventional activity of transmitting data (see MPEP 2106.05(d) example i in computer functions).
and updating, based on the updated policy, the existing Al model
At a high level of generality, this is an activity of using “the updated policy” as an “apply it” use (see MPEP 2106.05(f)). At said high level of generality, a generic recitation of “updating, based on the updated policy” the existing AI model does not incorporate the abstract idea into a practical invention and is seen as a variation of the phrase “apply it”.
In regards to Claim 11:
Step 2A Prong 1: Does the claim recite a law of nature, a natural phenomenon, or an abstract idea?
Yes, the claim does recite a(n) abstract idea.
Claim 11 recites the following abstract ideas:
Calculating, based on the Al model update period, that there is the difference
This limitation is directed towards the abstract idea of a mental process, or a concept performed in the human mind, including observation, evaluation, judgement or opinion (see MPEP 2106.04(a)(2) subsection 3). Here the limitation is seen as evaluation.
Step 2A Prong 2: Does the claim recite additional elements that integrate the exception into a practical application of the exception?
No, the application does not recite any additional elements that would integrate the abstract idea into a practical application.
Claim 11 recites the following additional elements:
obtaining, from a user, an Al model update period,
This limitation is directed towards the insignificant extra solution activity of mere data gathering (see MPEP § 2106.05(g)).
Step 2B: Does the claim as a whole amount to significantly more than the judicial exception?
No, the claim as a whole does not amount to significantly more than the judicial exception. All elements of the claim, viewed individually or wholistically, do not provide an inventive concept or otherwise significantly more than the abstract idea itself.
Claim 11 recites the following additional elements:
obtaining, from a user, an Al model update period,
This limitation is directed towards the insignificant extra solution activity of mere data gathering (see MPEP § 2106.05(g)). This is a well understood, routine, conventional activity of transmitting data (see MPEP 2106.05(d) example i in computer functions).
In regards to Claim 12:
Step 1: Is the claim directed towards a process, machine, manufacture, or composition of matter?
Yes, the claim is directed towards a device, so a machine.
Step 2A Prong 1: Does the claim recite a law of nature, a natural phenomenon, or an abstract idea?
Yes, the claim does recite a(n) abstract idea.
Claim 12 recites the same abstract ideas as analogous claim 1.
Step 2A Prong 2: Does the claim recite additional elements that integrate the exception into a practical application of the exception?
No, the application does not recite any additional elements that would integrate the abstract idea into a practical application.
Claim 12 recites the same additional elements as claim 1 aside what is listed below:
A computing device, comprising: a memory configured to store instructions; and a processor coupled to the memory and configured to execute the instructions to
At a high level of generality, this is an activity of using a computing device as an “apply it” use (see MPEP 2106.05(f)). At said high level of generality, a computing device appears to be an implementation of the abstract idea on a computer, so merely using a computer as a tool to perform the abstract idea.
Step 2B: Does the claim as a whole amount to significantly more than the judicial exception?
No, the claim as a whole does not amount to significantly more than the judicial exception. All elements of the claim, viewed individually or wholistically, do not provide an inventive concept or otherwise significantly more than the abstract idea itself.
Claim 12 recites the same additional elements as claim 1 aside what is listed below:
A computing device, comprising: a memory configured to store instructions; and a processor coupled to the memory and configured to execute the instructions to
At a high level of generality, this is an activity of using a computing device as an “apply it” use (see MPEP 2106.05(f)). At said high level of generality, a computing device appears to be an implementation of the abstract idea on a computer, so merely using a computer as a tool to perform the abstract idea.
In regards to Claim 13:
Step 2A Prong 1: Does the claim recite a law of nature, a natural phenomenon, or an abstract idea?
Yes, the claim does recite a(n) abstract idea.
Claim 13 recites the same abstract ideas as analogous claim 2.
Step 2A Prong 2: Does the claim recite additional elements that integrate the exception into a practical application of the exception?
No, the application does not recite any additional elements that would integrate the abstract idea into a practical application.
Claim 13 recites the same additional elements as claim 2.
Step 2B: Does the claim as a whole amount to significantly more than the judicial exception?
No, the claim as a whole does not amount to significantly more than the judicial exception. All elements of the claim, viewed individually or wholistically, do not provide an inventive concept or otherwise significantly more than the abstract idea itself.
Claim 13 recites the same additional elements as analogous claim 2.
In regards to Claim 14:
Step 2A Prong 2: Does the claim recite additional elements that integrate the exception into a practical application of the exception?
No, the application does not recite any additional elements that would integrate the abstract idea into a practical application.
Claim 14 recites the same additional elements as claim 3.
Step 2B: Does the claim as a whole amount to significantly more than the judicial exception?
No, the claim as a whole does not amount to significantly more than the judicial exception. All elements of the claim, viewed individually or wholistically, do not provide an inventive concept or otherwise significantly more than the abstract idea itself.
Claim 14 recites the same additional elements as analogous claim 3.
In regards to Claim 16:
Step 2A Prong 1: Does the claim recite a law of nature, a natural phenomenon, or an abstract idea?
Yes, the claim does recite a(n) abstract idea.
Claim 16 recites the same abstract ideas as analogous claim 5.
In regards to Claim 17:
Step 2A Prong 2: Does the claim recite additional elements that integrate the exception into a practical application of the exception?
No, the application does not recite any additional elements that would integrate the abstract idea into a practical application.
Claim 17 recites the same additional elements as claim 6.
Step 2B: Does the claim as a whole amount to significantly more than the judicial exception?
No, the claim as a whole does not amount to significantly more than the judicial exception. All elements of the claim, viewed individually or wholistically, do not provide an inventive concept or otherwise significantly more than the abstract idea itself.
Claim 17 recites the same additional elements as analogous claim 6.
In regards to Claim 18:
Step 2A Prong 2: Does the claim recite additional elements that integrate the exception into a practical application of the exception?
No, the application does not recite any additional elements that would integrate the abstract idea into a practical application.
Claim 18 recites the same additional elements as claim 8.
Step 2B: Does the claim as a whole amount to significantly more than the judicial exception?
No, the claim as a whole does not amount to significantly more than the judicial exception. All elements of the claim, viewed individually or wholistically, do not provide an inventive concept or otherwise significantly more than the abstract idea itself.
Claim 18 recites the same additional elements as analogous claim 8.
In regards to Claim 19:
Step 2A Prong 2: Does the claim recite additional elements that integrate the exception into a practical application of the exception?
No, the application does not recite any additional elements that would integrate the abstract idea into a practical application.
Claim 19 recites the same additional elements as claim 10.
Step 2B: Does the claim as a whole amount to significantly more than the judicial exception?
No, the claim as a whole does not amount to significantly more than the judicial exception. All elements of the claim, viewed individually or wholistically, do not provide an inventive concept or otherwise significantly more than the abstract idea itself.
Claim 19 recites the same additional elements as analogous claim 10.
In regards to Claim 20:
Step 2A Prong 1: Does the claim recite a law of nature, a natural phenomenon, or an abstract idea?
Yes, the claim does recite a(n) abstract idea.
Claim 20 recites the same abstract ideas as analogous claim 11.
Step 2A Prong 2: Does the claim recite additional elements that integrate the exception into a practical application of the exception?
No, the application does not recite any additional elements that would integrate the abstract idea into a practical application.
Claim 20 recites the same additional elements as claim 11.
Step 2B: Does the claim as a whole amount to significantly more than the judicial exception?
No, the claim as a whole does not amount to significantly more than the judicial exception. All elements of the claim, viewed individually or wholistically, do not provide an inventive concept or otherwise significantly more than the abstract idea itself.
Claim 20 recites the same additional elements as analogous claim 11.
In regards to Claim 21:
Step 2A Prong 1: Does the claim recite a law of nature, a natural phenomenon, or an abstract idea?
Yes, the claim does recite a(n) abstract idea.
Claim 21 recites the following abstract ideas:
Wherein the offline condition comprises the difference between the first data distribution of the inference data set and the second data distribution of the training data set being greater than a preset value
This limitation is directed towards the continuation abstract idea of a mental process, or a concept performed in the human mind, including observation, evaluation, judgement or opinion (see MPEP 2106.04(a)(2) subsection 3). Here the limitation is seen as a continuation of evaluation and judgement.
In regards to Claim 22:
Step 2A Prong 1: Does the claim recite a law of nature, a natural phenomenon, or an abstract idea?
Yes, the claim does recite a(n) abstract idea.
Claim 22 recites the same abstract ideas as analogous claim 1.
Step 2A Prong 2: Does the claim recite additional elements that integrate the exception into a practical application of the exception?
No, the application does not recite any additional elements that would integrate the abstract idea into a practical application.
Claim 22 recites the same additional elements as claim 1.
Step 2B: Does the claim as a whole amount to significantly more than the judicial exception?
No, the claim as a whole does not amount to significantly more than the judicial exception. All elements of the claim, viewed individually or wholistically, do not provide an inventive concept or otherwise significantly more than the abstract idea itself.
Claim 22 recites the same additional elements as analogous claim 1.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claims 1, 2, 21 is/are rejected under 35 U.S.C. 103 as being unpatentable over Tanimoto et al (US 20180082185 A1), referred to as Tanimoto in this document, and further in combination with Gottschlitch et al (US 20180129970 A1), referred to as Gottschlitch in this document.
Regarding Claim 1:
Tanimoto teaches:
obtaining an inference data set comprising inference data, wherein the inference data is for inputting into an existing artificial intelligence (AI) model for performing inference
[Tanimoto 0003]: “The apparatus described in PTL 1 sequentially updates energy demand prediction models whenever a predetermined period has passed, using data acquired a day ago, [obtaining an inference data set comprising inference data, wherein the inference data is for inputting into an existing artificial intelligence (AI) model for performing inference] data acquired an hour ago, or data acquired a minute ago.”
training, by using the inference data set and in response to a difference between a first data distribution of the inference data set and a second data distribution of a training data set meeting an offline update condition, the existing AI model offline to obtain an updated AI model
training, by using the inference data set and in response to the difference between the first data distribution of the inference data set and the second data distribution of the training data set not meeting the offline update condition, the existing AI model online to obtain the updated AI model
[Tanimoto 0027]: “The predictive model evaluation unit 13 determines whether or not to update the pre-relearning predictive model with the relearned predictive model [training, by using the inference data set and in response to a difference…, the existing AI model offline to obtain an updated AI model][ [training, by using the inference data set and in response to the difference…, the existing AI model online to obtain the updated AI model]]. In detail, the predictive model evaluation unit 13 extracts an update target predictive model, based on a rule (hereafter referred to as “update evaluation rule”) for determining whether or not to actually update the predictive model with the relearned predictive model. The update evaluation rule is a rule prescribing the status of change between the predictive model before update and the predictive model after update.”
[Tanimoto 0028]: “The status of change prescribed by the update evaluation rule may be any status of change. In this exemplary embodiment, the predictive model evaluation unit 13 focuses on the closeness in property of the predictive model, to determine the status of change between the predictive model before update and the predictive model after update [difference between a first data distribution of the inference data set and a second data distribution of a training data set meeting an offline update condition] [difference between a first data distribution of the inference data set and a second data distribution of a training data set not meeting an offline update condition]. In other words, the predictive model evaluation unit 13 evaluates the closeness in property between the relearned predictive model and the pre-relearning predictive model.”
Where the training data set is shown by the model that was created using the training data (such as the predictive model before update as whatever data the model was trained on would be training data). The difference is distribution is found by the difference in closeness or accuracy of the models, as accuracy is an indicator of distribution shift according to [current application 0003]: “Therefore, although the constructed AI model has a specific generalization capability, when there is a large difference between data distribution in a scenario in which the AI model is used and distribution of training data of the AI model, performance of the AI model is affected, and precision is reduced.”
[Tanimoto 0029]: “The closeness in property of the predictive model means at least the closeness in prediction result or the structural closeness of the predictive model. Thus, in this exemplary embodiment, the predictive model is kept from changing greatly by evaluating the change in property of the predictive model, in addition to improving the accuracy of the predictive model.”
Tanimoto does not explicitly teach:
training, by using the inference data set and in response to a difference between a first data distribution of the inference data set and a second data distribution of a training data set meeting an offline update condition, the existing AI model offline to obtain an updated AI model
training, by using the inference data set and in response to the difference between the first data distribution of the inference data set and the second data distribution of the training data set not meeting the offline update condition, the existing AI model online to obtain the updated AI model
Gottschlitch teaches:
Initial support and context in [Gottschlitch 0068]: "Example 1 is a machine-learning decision system (MLDS) apparatus, comprising: an online decision system to produce a first time slice-specific decision output corresponding to a first time slice based on one or more situational inputs received in the first time slice; and an offline decision system to produce a second time slice-specific decision output corresponding to the first time slice based on one or more situational inputs received in the first time slice and in a plurality of subsequent time slices occurring after the first time slice; and an online training engine to conduct negative-reinforcement training of the online decision system in response to a nonconvergence between the first and the second time slice-specific decision outputs."
training, by using the inference data set and in response to a difference between a first data distribution of the inference data set and a second data distribution of a training data set meeting an offline update condition, the existing AI model offline to obtain an updated AI model
[Gottschlitch 0070] In Example 3, the subject matter of any one or more of Examples 1-2 optionally include an offline training engine [training, by using the inference data set and in response to a difference… meeting an offline update condition, the existing AI model offline to obtain an updated AI model] to conduct unsupervised positive-reinforcement training of the offline decision system in response to a nonconvergence between the first and the second time slice-specific decision outputs.
training, by using the inference data set and in response to the difference between the first data distribution of the inference data set and the second data distribution of the training data set not meeting the offline update condition, the existing AI model online to obtain the updated AI model
[Gottschlitch 0071] In Example 4, the subject matter of any one or more of Examples 1-3 optionally include wherein the online training engine [training, by using the inference data set and in response to the difference… not meeting the offline update condition, the existing AI model online to obtain the updated AI model] is further to conduct positive-reinforcement training of the online decision system in response to a convergence between the first and the second time slice-specific decision outputs.
In accordance with the remarks to provide clarification for the rejection, figure 8 of Gottschlitch shows the comparison of prediction for determining of whether to train online or offline. This is seen as matching the indication in the current application figure 12 showing that a comparison of prediction is used to determine the difference (shown by “Prediction error”). Thus Gottschlitch is showing a difference being used to indicate whether to train offline or online as the comparison of prediction is seen as measuring the difference.
One of ordinary skill in the art, prior to the effective filing date, would have been motivated to combine Tanimoto and Gottschlitch. Tanimoto and Gottschlitch are in the same field of endeavor of machine learning. One of ordinary skill in the art would have been motivated to combine in order to improve machine learning to account for issues of realtime learning or learning on the fly ([Gottschlitch 0003]: “In many applications, such as complex data centers, self-driving vehicles, security systems, medical treatment systems, transaction systems, and the like, the detection of anomalies is expected to occur in real time, i.e., with short latency from the present. This expectation complicates the training of decision systems. Conventional training methods including supervised machine learning, unsupervised learning, and reinforcement learning, tend to be either too slow (i.e., not suitable for real-time, or online learning on the fly), or lacking accuracy when assessing predictions based on past or present results. A practical, computationally-efficient solution is needed for decision systems for anomaly-detection, and various other applications.”)
Regarding Claim 2:
The method of claim 1 is taught by Tanimoto and Gottschlitch.
Tanimoto teaches:
wherein the existing AI model is deployed on an inference platform, and wherein the method further comprises: comparing a first inference precision of the updated Al model with a second inference precision of the existing AI model;
deploying, in response to the first inference precision being higher than the second inference precision, the updated Al model on the inference platform to perform inference in place of the existing Al model
[Tanimoto 0027]: “The predictive model evaluation unit 13 determines whether or not to update the pre-relearning predictive model [wherein the existing AI model is deployed on an inference platform as the model is predictive thus performing inference] with the relearned predictive model [deploying, in response to the first inference precision being higher than the second inference precision, the updated Al model on the inference platform to perform inference in place of the existing Al model where the precision being higher is shown below]. In detail, the predictive model evaluation unit 13 extracts an update target predictive model, based on a rule (hereafter referred to as “update evaluation rule”) for determining whether or not to actually update the predictive model with the relearned predictive model. The update evaluation rule is a rule prescribing the status of change between the predictive model before update and the predictive model after update.”
[Tanimoto 0028]: “The status of change prescribed by the update evaluation rule may be any status of change. In this exemplary embodiment, the predictive model evaluation unit 13 focuses on the closeness in property of the predictive model, to determine the status of change between the predictive model before update and the predictive model after update. In other words, the predictive model evaluation unit 13 evaluates the closeness in property between the relearned predictive model and the pre-relearning predictive model.”
[Tanimoto 0029]: “The closeness in property of the predictive model means at least the closeness in prediction result [and wherein the method further comprises: comparing a first inference precision of the updated Al model with a second inference precision of the existing AI model;] or the structural closeness of the predictive model. Thus, in this exemplary embodiment, the predictive model is kept from changing greatly by evaluating the change in property of the predictive model, [deploying, in response to the first inference precision being higher than the second inference precision,…] in addition to improving the accuracy of the predictive model.”
Regarding Claim 21:
The method of claim 1 is taught by Tanimoto and Gottschlitch.
Tanimoto does not explicitly teach:
wherein the offline condition comprises the difference between the first data distribution of the inference data set and the second data distribution of the training data set being greater than a preset value
Gottschlitch teaches:
wherein the offline condition comprises the difference between the first data distribution of the inference data set and the second data distribution of the training data set being greater than a preset value
[Gottschlitch 0070] In Example 3, the subject matter of any one or more of Examples 1-2 optionally include an offline training engine [wherein the offline condition comprises the difference between the first data distribution of the inference data set and the second data distribution of the training data set being greater than a preset value] to conduct unsupervised positive-reinforcement training of the offline decision system in response to a nonconvergence between the first and the second time slice-specific decision outputs.
Following the teachings presented around training offline in claim 1, Gottschlitch teaches training offline based on the difference between two data distributions (as noted in further clarity in claim 1 around Figure 8 of Gottschlitch). Here the “being greater than a preset value” is taught by the nonconvergence as nonconvergence would be a difference greater than 0 (where the value for convergence is assumed to be 0 for the sake of explanation).
The motivation to combine with Gottschlitch is the same as in claim 1.
Claims 3, 6, 7, 11-14, 17, 20, 22 is/are rejected under 35 U.S.C. 103 as being unpatentable over Tanimoto et al (US 20180082185 A1), referred to as Tanimoto in this document, and further in combination with Gottschlitch et al (US 20180129970 A1), referred to as Gottschlitch in this document, and further in combination with Ikeda et al (US 20200334578 A1), referred to as Ikeda in this document.
Regarding Claim 3:
The method of claim 2 is taught by Tanimoto and Gottschlitch.
Tanimoto teaches:
wherein before deploying the updated Al model, the method further comprises: displaying, by using a display interface, the first inference precision and the second inference precision; and receiving, from a user in response to displaying the first inference precision and the second inference precision, an update instruction for the existing Al model
[Tanimoto 0050]: “The predictive model updating unit 14 updates the pre-relearning predictive model with the relearned predictive model, in the case where the closeness in property between both predictive models evaluated by the predictive model evaluation unit 13 meets the condition prescribed by the update evaluation rule. The update evaluation rule prescribes the closeness that allows updating the predictive model, depending on the evaluation. The predictive model updating unit 14 may alert the user [wherein before deploying the updated Al model, the method further comprises: displaying, by using a display interface, the first inference precision and the second inference precision], instead of automatically updating the predictive model [and receiving, from a user in response to displaying the first inference precision and the second inference precision, an update instruction for the existing Al model]. Any alerting method may be used, such as display on a screen or notification by mail.”
[Tanimoto 0051]: “The result output unit 15 outputs the relearning result by the predictive model relearning unit 12 and/or the update result by the predictive model updating unit 14. The result output unit 15 may display the relearning result and/or the update result on a display device (not depicted).”
Tanimoto notes that display methods can be used to show things such as relearning or update results, and that a user may be alerted instead of performing an update based on the closeness evaluation automatically. Thus the displaying of values like inference precision for the user to decide upon the update is seen as taught by Tanimoto.
Tanimoto does not explicitly teach:
from a user in response to displaying
Ikeda teaches:
from a user in response to displaying
Computer elements, such as processor, memory, and hardware for user input are taught in [Ikeda 0038]: “The memory device 153 reads the program from the auxiliary storage device 152 and stores the program in response to an instruction to start the program. The CPU 154 achieves a function related to the model training apparatus 100 in accordance with the program stored in the memory device 153. The interface device 155 is used as an interface for connecting to a network. The display device 156 displays a Graphical User Interface (GUI) or the like implemented by the program. The input device 157 includes a keyboard and a mouse, a button, a touch panel, and the like, and is used to input various operating instructions [from a user in response to displaying as Ikeda teaches the elements required for the user to provide feedback]. The display device 156 may not be provided.”
One of ordinary skill in the art, prior the effective filing date, would have been motivated to combine Tanimoto and Ikeda. Tanimoto and Ikeda are in the same field of endeavor of machine learning. One of ordinary skill in the art would have been motivated to combine Tanimoto and Ikeda in order to incorporate user feedback to improve the system, such as for getting a correct label ([Ikeda 0068]: “As a method of generating a correct label, for example, there is a method of classifying data into classes to which respective data belong from a log at times of obtaining respective data based on a correspondence relation between a log and each class, which is predetermined based on the log generated from an object to be monitored. Other methods include a method of inputting a result classified by another classification system operating in parallel, a method of generating a result based on a report from a user currently using a system (for example, classifying the quality in use into “high”, “medium”, and “low”, and receiving feedback on a result of classifying the quality from the user), and a method of inputting a result visually determined by a system operator.”), and working with labeling data to fix drifting ([Ikeda 0007 under Problem to be Solved by the Invention]: “For example, when the data is data such as traffic amount, the traffic amount tends to increase in an overall network, input data to the anomaly detection algorithm gradually increases, and a simple increase in the traffic amount may be detected as an anomaly. Additionally, there is a problem to be solved that when over-detection that determines normal data as the anomaly and non-detection that overlooks abnormal data as normal data, occur, similar over-detection and non-detection repeatedly occur.”).
Regarding Claim 6:
The method of claim 1 is taught by Tanimoto and Gottschlitch.
Tanimoto does not explicitly teach:
wherein training the existing Al model comprises: constructing, based on the inference data set, a target data set; and updating, by using the target data set, the existing Al model
Ikeda teaches:
wherein training the existing Al model comprises: constructing, based on the inference data set, a target data set; and updating, by using the target data set, the existing Al model
[Ikeda 0056]: “According to a disclosed technique, an apparatus for training a model including a storage unit configured to store a parameter of the model trained by using a training data set, and the training data set, a detector configured to use the model to determine whether an anomaly is present in a test data set and store a determined result and the test data set in the storage device [wherein updating the existing Al model comprises: constructing, based on the inference data set, a target data set where the result is seen as a label labeling the test data (thus creating a target data set)], and a retraining unit configured to retrain the model [and updating, by using the target data set, the existing Al model] by using the determined result, the test data set, and the training data set, is provided”
The motivation to combine with Ikeda is the same motivation provided in claim 3 to combine with Ikeda.
Regarding Claim 7:
The method of claim 6 is taught by Tanimoto, Gottschlitch, and Ikeda.
Ikeda teaches:
wherein constructing the target data set comprises: obtaining, from the inference data set, target data that meets a sample condition; displaying, by using a display interface, the target data; obtaining, from a user, a result of labeling the target data; and constructing, based on the target data and the result, the target data set
[Ikeda 0056]: “For example, as a method of generating a correct label [wherein constructing the target data set comprises: obtaining, from the inference data set, target data that meets a sample condition as generating a label for data can be seen as creating a target data set utilizing unlabeled data (that would be akin to the inference data set in this case) and “a sample condition” is seen as met as any condition of the data could be considered meeting a condition][ and constructing, based on the target data and the result, the target data set], there is a method that determines data is “normal” when the data has been obtained in a normal operation and determines the data is abnormal when a predetermined special operation has been performed, based on a log generated from an object to be monitored. Other methods include a method inputting a result determined by another monitoring system that is operated in parallel, a method generating a label based on a report indicating the anomaly from a user currently using a system, and a method inputting a result visually [displaying, by using a display interface, the target data] determined by a system operator [obtaining, from a user, a result of labeling the target data].”
The motivation to combine with Ikeda is the same motivation to combine with Ikeda in claim 6.
Regarding Claim 11:
The method of claim 1 is taught by Tanimoto and Gottschlitch.
Tanimoto teaches:
obtaining, from a user,
Tanimoto notes user interaction in [Tanimoto 0050]: “The predictive model updating unit 14 may alert the user, instead of automatically updating the predictive model. Any alerting method may be used, such as display on a screen or notification by mail.”, but Tanimoto does not explicitly note elements for the user input.
an Al model update period and calculating, based on the AI model update period, the difference between the first data distribution of the inference data set and the second data distribution of the training data set
[Tanimoto 0022]: “The predictive model update determination unit 11 determines a predictive model of an update candidate. In detail, the predictive model update determination unit 11 extracts a relearning target predictive model as an update candidate from a plurality of predictive models, based on a rule (hereafter referred to as “relearning rule”) for determining whether or not to relearn the predictive model. The relearning rule is a rule prescribing, based on a predetermined evaluation index, whether or not the predictive model needs to be relearned.
[Tanimoto 0023]: “The evaluation index used in the relearning rule may be any index. Examples of the evaluation index include the period from the previous learning of the predictive model, the period [an Al model update period and calculating, based on the AI model update period, the difference between the first data distribution of the inference data set and the second data distribution of the training data set where the specifics of performing the difference determination are taught in claim 1, as Tanimoto is teaching the update period as a trigger] from the previous update of the predictive model, the amount of increase of learning data, the degree of accuracy degradation over time, the change of the number of samples, and the computational resources. The evaluation index is, however, not limited to such, and any index that can be used to determine whether or not to update the predictive model may be used. The evaluation index is also not limited to data calculated from the prediction result.”
Tanimoto does not explicitly teach:
obtaining, from a user,
Ikeda teaches:
obtaining, from a user,
Computer elements, such as processor, memory, and hardware for user input are taught in [Ikeda 0038]: “The memory device 153 reads the program from the auxiliary storage device 152 and stores the program in response to an instruction to start the program. The CPU 154 achieves a function related to the model training apparatus 100 in accordance with the program stored in the memory device 153. The interface device 155 is used as an interface for connecting to a network. The display device 156 displays a Graphical User Interface (GUI) or the like implemented by the program. The input device 157 includes a keyboard and a mouse, a button, a touch panel, and the like, and is used to input various operating instructions [obtaining, from a user,]. The display device 156 may not be provided.”
The motivation to combine with Ikeda is the same motivation as claim 6.
Regarding Claim 12:
This claim is analogous to claim 1, aside from the limitation below:
Tanimoto does not explicitly teach:
A computing device, comprising:, a memory configured to store instructions; and a processor coupled to the memory and configured to execute the instructions to
Ikeda teaches:
A computing device, comprising:, a memory configured to store instructions; and a processor coupled to the memory and configured to execute the instructions to
[Ikeda 0038]: “The memory device [A computing device, comprising:, a memory configured to store instructions] 153 reads the program from the auxiliary storage device 152 and stores the program in response to an instruction to start the program. The CPU [and a processor coupled to the memory and configured to execute the instructions to] 154 achieves a function related to the model training apparatus 100 in accordance with the program stored in the memory device 153. The interface device 155 is used as an interface for connecting to a network. The display device 156 displays a Graphical User Interface (GUI) or the like implemented by the program. The input device 157 includes a keyboard and a mouse, a button, a touch panel, and the like, and is used to input various operating instructions. The display device 156 may not be provided.”
[Ikeda claim 8]: “A non-transitory computer-readable [non-transitory computer readable medium] recording medium having a program for causing a computer to perform each function of the apparatus for training the model claimed in claim 1.”
One of ordinary skill in the art, prior the effective filing date, would have been motivated to combine Tanimoto and Ikeda. Tanimoto and Ikeda are in the same field of endeavor of machine learning. One of ordinary skill in the art would have been motivated to combine Tanimoto and Ikeda in order to incorporate the invention into a physical device ([Ikeda 0038]: “The memory device 153 reads the program from the auxiliary storage device 152 and stores the program in response to an instruction to start the program. The CPU 154 achieves a function related to the model training apparatus 100 in accordance with the program stored in the memory device 153. The interface device 155 is used as an interface for connecting to a network. The display device 156 displays a Graphical User Interface (GUI) or the like implemented by the program. The input device 157 includes a keyboard and a mouse, a button, a touch panel, and the like, and is used to input various operating instructions. The display device 156 may not be provided.”).
Regarding Claim 13:
The device of claim 12 is taught by Tanimoto, Gottschlitch, and Ikeda.
This claim is analogous to claim 2.
Regarding Claim 14:
The device of claim 13 is taught by Tanimoto, Gottschlitch, and Ikeda.
This claim is analogous to claim 3.
Regarding Claim 17:
The device of claim 12 is taught by Tanimoto, Gottschlitch, and Ikeda.
This claim is analogous to claim 6.
Regarding Claim 20:
The device of claim 12 is taught by Tanimoto, Gottschlitch, and Ikeda.
This claim is analogous to claim 11.
Regarding Claim 22:
Claim 22 is analogous to claim 12.
Claims 5 is/are rejected under 35 U.S.C. 103 as being unpatentable over Tanimoto et al (US 20180082185 A1), referred to as Tanimoto in this document, and further in combination with Gottschlitch et al (US 20180129970 A1), referred to as Gottschlitch in this document, and further in combination with Namsoon et al (US 11829871 B2), referred to as Namsoon in this document.
Regarding Claim 5:
The method of claim 1 is taught by Tanimotto and Gottschlitch.
Tanimoto does not explicitly teach:
wherein training the existing AI model offline comprises: calculating, by using the difference, a parameter change amount of a first target part in the existing AI model; and calculating, based on a current parameter and the parameter change amount, a parameter of a second target part in the updated AI model
Namsoon teaches:
wherein training the existing AI model offline comprises: calculating, by using the difference, a parameter change amount of a first target part in the existing AI model; and calculating, based on a current parameter and the parameter change amount, a parameter of a second target part in the updated AI model
[Namsoon Column 3 Line 38]: “Embodiments of the present disclosure include validating a performance of a neural network trained using labeled training and validation data generated based on data collected by a device, including: determining proposed model parameters as potential updates to the neural network using the labeled validation data; performing a short-term validation on the proposed model parameters applied to the neural network based on the labeled validation data by comparing a first performance output and a second performance output; updating the currently-existing model parameters with the proposed model parameters when the second performance output outperforms the first performance output with respect to the labeled validation data [wherein training the existing AI model offline comprises: (where the online and offline teachings are taught in claim 1) calculating, by using the difference, a parameter change amount of a first target part in the existing AI model; where the first target part is the parameters of the currently-existing model], performing a long-term validation on the updated currently-existing model parameters applied to the neural network by determining a difference between original model parameters applied to the neural network and the updated currently-existing model parameters [and calculating, based on a current parameter and the parameter change amount, a parameter of a second target part in the updated AI model as the second target part is the parameters of the updated currently-existing model] applied to the neural network; and performing an operation when the difference between the original model parameters and the updated currently-existing model parameters lies within a threshold. The labeled validation data is derived from a same dataset collected by the device as the labeled training data. The first performance output is determined from applying the proposed model parameters to the neural network and the second performance output is determined from applying currently-existing model parameters to the neural network. The updated currently-existing model parameters corresponds to up-to-date model parameters.”
One of ordinary skill in the art, prior to the effective filing date would have been motivated to combine Tanimoto and Namsoon. Tanimoto and Namsoon are in the same field of endeavor of machine learning. One of ordinary skill in the art would have been motivated to combine Tanimoto and Namsoo in order to improve or keep model performance over time ([Namsoon Column 2 Line 17]: “It follows that a possible solution to overcome the challenges of acquiring large enough training data without manual labeling is to automatically generate labels for unlabeled data. But there is no guarantee whether the neural network will perform within accepted parameters over a longer time period and after a series of model parameter updates based on the automatically generated labeled training data. Therefore, a need arises for the process of validating the performance of a neural network after a series of model parameter updates.”)
Claims 8, 18 is/are rejected under 35 U.S.C. 103 as being unpatentable over Tanimoto et al (US 20180082185 A1), referred to as Tanimoto in this document, and further in combination with Gottschlitch et al (US 20180129970 A1), referred to as Gottschlitch in this document, and further in combination with Ikeda et al (US 20200334578 A1), referred to as Ikeda in this document, and further in combination with Bolon-Candedo et al (“Feature selection for high-dimensional data”), referred to as Bolon-Candedo in this document.
Regarding Claim 8:
The method of claim 6 is taught by Tanimoto and Ikeda.
Tanimoto does not explicitly teach:
wherein the target data set comprises unlabeled data and labeled data, and wherein updating the existing Al model comprises: optimizing, by using the unlabeled data, a feature extraction part in the existing Al model in an unsupervised manner to produce an optimized feature extraction part; and updating, based on the optimized feature extraction part and the labeled data, the existing Al model
Bolon-Canedo teaches:
wherein the target data set comprises unlabeled data and labeled data,
[Bolon-Candedo 5 Open challenges page 9]: “In conclusion, although feature selection is a field of machine learning that has been applied for decades, it is still in the spotlight due to the advent of Big Data and the appearance of new scenarios—not only related with massive volumes or stream data, but also with other aspects such as unbalanced classes, uncertain and partial labels [wherein the target data set comprises unlabeled data and labeled data], non-stationary distributions, etc.—which open new lines of research in which the use of feature selection is, perhaps, more necessary than ever.”
and wherein updating the existing Al model comprises: optimizing, by using the unlabeled data, a feature extraction part in the existing Al model in an unsupervised manner to produce an optimized feature extraction part; and updating, based on the optimized feature extraction part and the labeled data, the existing Al model
[Bolon-Candedo 3.1 What is feature selection? Page 4]: “The ultrahigh dimensionality of actual data sets not only incurs unbearable memory requirements and high computational cost in training [and updating, based on the optimized feature extraction part and the labeled data, the existing Al model], but also deteriorates the generalization ability of learning algorithms because of the “curse of dimensionality” issue. This term, coined by Richard Bellman in [4], indicates the difficulty of optimization by exhaustive enumeration on product spaces. Considering that a data set can be represented by a matrix where the rows are the recorded samples and the columns are the features, to tackle the “curse of dimensionality” issue, we can find “narrower” matrices that in some sense are close to the original. Since these narrower matrices have a smaller number of features, they can be used much more efficiently than the original matrix [wherein updating the existing Al model comprises: optimizing, by using the unlabeled data, a feature extraction part in the existing Al model in an unsupervised manner to produce an optimized feature extraction part;]. The process of finding these narrow matrices is called dimensionality reduction. There are two main techniques to achieve this dimensionality reduction: feature extraction and feature selection. Feature extraction consists of reducing the feature space by deriving new features transforming the existing ones; these new features are intended to be informative and non-redundant. On the other hand, feature selection (FS) is defined as the process of detecting relevant features and discarding irrelevant and redundant features with the goal of obtaining a subset of features that accurately describe a given problem with a minimum degradation of performance [29]. Both techniques are aimed at improving the performance of machine learning methods by using simpler models, probably gaining training speed.”
One of ordinary skill in the art, prior the effective filing date, would have been motivated to combine Tanimoto and Bolon-Candedo. Tanimoto and Bolon-Candedo are in the same field of endeavor of machine learning. One of ordinary skill in the art would have been motivated to combine Tanimoto and Bolon-Candedo in order to improve efficiency in training and generalization of the model ([Bolon-Candedo 3.1 What is feature selection? Page 4]: “The ultrahigh dimensionality of actual data sets not only incurs unbearable memory requirements and high computational cost in training, but also deteriorates the generalization ability of learning algorithms because of the “curse of dimensionality” issue. This term, coined by Richard Bellman in [4], indicates the difficulty of optimization by exhaustive enumeration on product spaces. Considering that a data set can be represented by a matrix where the rows are the recorded samples and the columns are the features, to tackle the “curse of dimensionality” issue, we can find “narrower” matrices that in some sense are close to the original. Since these narrower matrices have a smaller number of features, they can be used much more efficiently than the original matrix”).
Regarding Claim 18:
The device of claim 17 is taught by Tanimoto and Ikeda.
This claim is analogous to claim 8.
Claims 9, 10, 19 is/are rejected under 35 U.S.C. 103 as being unpatentable over Tanimoto et al (US 20180082185 A1), referred to as Tanimoto in this document, and further in combination with Gottschlitch et al (US 20180129970 A1), referred to as Gottschlitch in this document, and further in combination with Ikeda et al (US 20200334578 A1), referred to as Ikeda in this document, and further in combination with Ando et al (US 20210166679 A1), referred to as Ando in this document.
Regarding Claim 9:
The method of claim 6 is taught by Tanimoto and Ikeda.
Tanimoto does not explicitly teach:
wherein the target data set comprises unlabeled data and labeled data, and wherein updating the existing Al model comprises: labeling, by using the existing Al model, the unlabeled data to obtain a labeling result of the unlabeled data; and updating, based on the labeling result and the labeled data, the existing Al model
Ando teaches:
wherein the target data set comprises unlabeled data and labeled data, and wherein updating the existing Al model comprises: labeling, by using the existing Al model, the unlabeled data to obtain a labeling result of the unlabeled data; and updating, based on the labeling result and the labeled data, the existing Al model
[Ando 0011]: “To solve the above-described problems, a self-training data selection apparatus according to a first aspect of the present invention includes an estimation model storage configured to store an estimation model for estimating confidence for each of predetermined labels from each of feature amounts extracted from input data, learned using a plurality of the independent feature amounts extracted from data with a teacher label, a confidence estimating part configured to estimate confidence for each of the labels from the feature amounts extracted from data with no teacher label [wherein the target data set comprises unlabeled data and labeled data,] using the estimation model, and a data selecting part configured to, when one feature amount selected from the feature amounts is set as a feature amount to be learned, the confidence for each label obtained from the data with no teacher label exceeds all confidence thresholds which are set in advance for each of the feature amounts for the feature amount to be learned, and labels for which confidence exceeds the confidence thresholds are the same in all feature amounts, add a label corresponding to the confidence which exceeds all the confidence thresholds to the data with no teacher label as a teacher label [and wherein updating the existing Al model comprises: labeling, by using the existing Al model, the unlabeled data to obtain a labeling result of the unlabeled data] to select the data as self-training [and updating, based on the labeling result and the labeled data, the existing Al model] data of the feature amount to be learned, and the confidence thresholds are set higher for a feature amount which is not to be learned than for the feature amount to be learned.”
One of ordinary skill in the art, prior to the effective filing date, would have been motivated to combine Tanimoto and Ando. Tanimoto and Ando are in the same field of endeavor of machine learning. One of ordinary skill in the art would have been motivated to combine in order to utilize data that does not contain a label ([Ando 0010]: “In view of such technical problems, an object of the present invention is to effectively self-train an estimation model by utilizing a large amount of data with no teacher label.”).
Regarding Claim 10:
The method of claim 6 is taught by Tanimoto and Ikeda.
Tanimoto does not explicitly teach:
wherein updating the existing Al model comprises: obtaining, based on a data characteristic of data in the target data set, a policy for updating the existing Al model to obtain an updated policy; and updating, based on the updated policy, the existing Al model
Ando teaches:
wherein updating the existing Al model comprises: obtaining, based on a data characteristic of data in the target data set, a policy for updating the existing Al model to obtain an updated policy; and updating, based on the updated policy, the existing Al model
[Ando 0005]: “Examples of a typical semi-supervised learning method can include self-training [a policy for updating the existing Al model to obtain an updated policy where policy is interpreted as a method and self-training is seen as the updated policy as self-training is a method/policy derived from using semi-supervised learning method. The limitation is interpreted as seeking a method to use for training.][and updating, based on the updated policy, the existing Al model as training/learning is seen as updating] (see Non-patent literature 3). Self-training is a method in which a label of unsupervised data is estimated using an estimation model learned from a few pieces of data with teacher labels, and the estimated label is relearned as a teacher label [wherein updating the existing Al model comprises: obtaining, based on a data characteristic of data in the target data set where the characteristic could be seen as the label or a property of the data]. At this time, only utterance with high confidence of the teacher label (such as, for example, utterance for which a posterior probability of a certain teacher label is equal to or higher than 90%) is learned.”
The motivation to combine with Ando in claim 10 is the same as the motivation to combine with Ando in claim 9, as the method/policy is for providing the benefit noted in the motivation for claim 9.
Regarding Claim 19:
The device of claim 17 is taught by Tanimoto and Ikeda.
This claim is analogous to claim 10.
Claims 16 is/are rejected under 35 U.S.C. 103 as being unpatentable over Tanimoto et al (US 20180082185 A1), referred to as Tanimoto in this document, and further in combination with Ikeda et al (US 20200334578 A1), referred to as Ikeda in this document, and further in combination with Gottschlitch et al (US 20180129970 A1), referred to as Gottschlitch in this document, and further in combination with Namsoon et al (US 11829871 B2), referred to as Namsoon in this document.
Regarding Claim 16:
The device of claim 12 is taught by Tanimoto, Ikeda, and Gottschlitch.
This claim is analogous to claim 5.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Nazari et al (US 20190102676 A1) is relevant art that teaches aspects of learning online and offline, policies for training, user input, data sets, and accuracy as a metric to determine if a model is good.
Subbaswamy et al (“From development to deployment: dataset shift, causality, and shift-stable models in health AI”) is relevant art that covers aspects related to detecting datashifts in inference data, updating or replacing a model, and strategies for updating or replacing a model (like a policy).
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). 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 CHRISTOPHER D DEVORE whose telephone number is (703)756-1234. The examiner can normally be reached Monday-Friday 7:30 am - 5 pm EST.
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, Michael J Huntley can be reached at (303) 297-4307. 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.D.D./Examiner, Art Unit 2129
/MICHAEL J HUNTLEY/Supervisory Patent Examiner, Art Unit 2129