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 .
Specification
The disclosure is objected to because of the following informalities: ¶[0099] designates reference number 810 as the “first determination unit” for determining the set of first training data; however, ¶[0100] and ¶[0103] consistently designate reference 810 as the “1st ML system” (i.e., the first machine-learning system for building the regularized machine-learning model), with the “determination unit” separately assigned reference number 812 in ¶[0103] and Fig. 8. The assignment of reference 810 to “first determination unit” in ¶[0099] appears to be an error; the correct reference for the determination unit for determining the set of first training data is 812.
Appropriate correction is required.
Claim Objections
Claims 4 and 14 are objected to because of the following informalities: each claim recites “a Ridge regression algorithm an Elastic Net regression algorithm” with a missing comma between “Ridge regression algorithm” and “an Elastic Net regression algorithm.” The correct language should read “a Ridge regression algorithm, an Elastic Net regression algorithm.”
Appropriate correction is required.
Claim Rejections - 35 USC § 112
The following is a quotation of the first paragraph of 35 U.S.C. 112(a):
(a) IN GENERAL.—The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor or joint inventor of carrying out the invention.
The following is a quotation of the first paragraph of pre-AIA 35 U.S.C. 112:
The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor of carrying out his invention.
Written Description
Claims 1-20 are rejected under 35 U.S.C. 112(a) or 35 U.S.C. 112 (pre-AIA ), first paragraph, as failing to comply with the written description requirement. The claim(s) contains subject matter which was not described in the specification in such a way as to reasonably convey to one skilled in the relevant art that the inventor or a joint inventor, at the time the application was filed, had possession of the claimed invention.
Per claim 1: Claim 1 recites "training of a first machine-learning system for building a regularized machine-learning model, using the set of first training data and respective measured sampling time-series data as target data, thereby determining a subset of the set of first training data, the subset relating to a set of most influential time lag values…". The claim thus requires that the regularized machine-learning model, by virtue of its training, reduce the first training data to a proper subset corresponding to the most influential time lag values.
The specification conveys possession of this subset-determining functionality only for a LASSO (L1) regularized model. The specification explains that LASSO "encourage[s] sparse solutions, where some coefficients may be forced to be exactly ZERO," which renders it "particularly useful for feature vector selection because it may automatically identify and discard irrelevant or redundant variables" (¶[0035]; see also ¶[0061] and ¶[0079] ("'unimportant' time lag values are eliminated" "as a consequence of the characteristics of the LASSO regression training algorithm")).
The term "regularized machine-learning model," however, is not limited to LASSO. The specification defines regularization broadly (¶[0060]: "[t]he two main types of regularization techniques are Ridge Regularization and LASSO Regularization") and, critically, states that "[i]n regularization, one keeps the same number of feature vectors but reduces the magnitude of the related coefficients" (¶[0063]). A regularized model of the Ridge type, as the specification itself describes it, therefore does not eliminate any feature vector and cannot produce the claimed reduced "subset ... relating to a set of most influential time lag values." The specification discloses no algorithm or mechanism by which a regularized machine-learning model other than LASSO accomplishes the claimed determination of a subset. Disclosure of a single species (LASSO), supported by a particular structural mechanism (L1 sparsity), does not convey possession of a genus claimed by the functional result of subset-determination, where other members of the genus are described as not performing feature elimination. See MPEP § 2163.
Per claim 4: Claim 4 affirmatively recites that the regularized machine-learning model is "selected from the group consisting of a Least Absolute Shrinkage and Selection Operator regression algorithm, a Ridge regression algorithm[,] an Elastic Net regression algorithm, and a tree-based machine-learning model." The specification provides no description of how a Ridge regression model (which retains all feature vectors, ¶[0063]) or a tree-based machine-learning model functions as a regularized model that "determin[es] a subset of the set of first training data ... relating to a set of most influential time lag values" as required by parent claim 1. Merely naming members of a Markush group does not convey possession of each member performing the claimed subset-determination function. See MPEP § 2163.
Per claims 11 and 20: Independent claim 11 (system) and independent claim 20 (computer program product) recite the identical "regularized machine-learning model ... thereby determining a subset of the set of first training data" limitation and are rejected for the same reasons as claim 1. Reciting generic computer hardware (claim 11: "one or more processors," "computer-readable memories," "tangible storage medium"; see ¶[0099]-[0102], ¶[0106]-[0121]) or generic program instructions on a storage medium (claim 20; see ¶[0013], ¶[0104]-[0105]) does not supply possession of the full scope of the recited regularized-model genus. Claim 14 recites the same Markush group as claim 4 and is rejected for the reasons given for claim 4.
Per claims 2-3, 5-10, 12-13, and 15-19: These dependent claims do not remedy the deficiency of the independent claim from which they depend and are therefore rejected under 35 U.S.C. 112(a) for the same reasons. See MPEP § 2163.
Enablement
Claims 1, 4, 11, 14, and 20 are rejected under 35 U.S.C. 112(a) or 35 U.S.C. 112 (pre-AIA ), first paragraph, as failing to comply with the enablement requirement. The claim(s) contains subject matter pertaining to the “regularized machine-learning model” which was not described in the specification in such a way as to enable one skilled in the art to which it pertains, or with which it is most nearly connected, to make and/or use the invention commensurate in scope with these claims.
The factors to be considered in determining whether a disclosure meets the enablement requirement of 35 U.S.C. 112(a) have been described in In re Wands, 858 F.2d 731, 737 (Fed. Cir. 1988). See MPEP § 2164.01.
(1) Breadth of the claims: Broad. The independent claims encompass any "regularized machine-learning model" that, through training, determines a subset of the first training data relating to the most influential time lag values
(2) Nature of the invention: A computer-implemented technique that uses a regularized machine-learning model to reduce/select time-lag features for downstream time-series forecasting.
(3) State of the prior art: Regularized regression (LASSO, Ridge, Elastic Net) and tree-based models are individually well known; however, feature elimination yielding a reduced feature subset is a property of L1/sparsity-inducing penalties, not of regularization generally.
(4) Level of one of ordinary skill: High (e.g., an advanced degree or equivalent experience in machine learning or statistics).
(5) Level of predictability in the art: Although software/ML is generally predictable, whether a given regularized model produces a reduced feature subset is not predictable across the claimed genus. The specification states that regularization "keeps the same number of feature vectors" (¶[0063]), confirming that Ridge-type members do not eliminate features.
(6) Amount of direction or guidance provided: The specification provides direction sufficient to practice the subset-determination only for LASSO (¶[0035], [0061], [0079]). It provides no parameters, procedure, or worked guidance for achieving the claimed subset-determination using Ridge, Elastic Net, or tree-based models.
(7) Existence of working examples: The disclosure is prophetic and LASSO-centric. No example demonstrates a non-LASSO regularized model producing the claimed reduced subset of most influential time lag values.
(8) Quantity of experimentation needed: Undue. To practice the full scope, a PHOSITA would have to independently devise how a Ridge or tree-based "regularized machine-learning model" eliminates lag features to form a reduced subset — directly contrary to the specification's statement that regularization retains all feature vectors.
Considering the Wands factors in totality, the specification enables a PHOSITA to practice the claimed subset-determination using a LASSO model but does not enable the full scope of the claimed "regularized machine-learning model." The enabling disclosure is therefore narrower than the claimed genus. See MPEP § 2164.01.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 1-20 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Claims 1, 11, and 20 each recite (1) "determining that an element of the set of second training data is significant for training of the first time-series machine-learning model," and (2) "determining, after the determining that an element of the second training data is significant, that the set of second training data is complete."
Regarding the term "significant": Neither the claim nor the specification provides an objective standard, criterion, threshold, or metric for determining whether an element of the second training data is or is not "significant" for purposes of this limitation. The specification at ¶[0049] describes one possible approach (deselecting an element if its exclusion results in a smaller or equal performance indicator value), and ¶[0065] states that the most influential time lag values "may be separated by threshold techniques," but no threshold, criterion, or test is defined in claims 1, 11, or 20. Different persons of ordinary skill in the art may apply different criteria to determine significance, rendering the scope of the limitation not reasonably certain. The dependent claims 5 and 15 add specificity for the determination of significance, but this does not cure the indefiniteness of the independent claims. See MPEP § 2173.05(b).
Regarding the term "complete": Neither the claim nor the specification provides an objective termination condition, criterion, or test for determining when the set of second training data is "complete." The specification at ¶[0071] states that this term is "used in the context of the forward optimization" but does not define completeness. ¶[0094] describes a stop condition (FPP < a predefined value, e.g., 0.7) but this criterion is not recited in claims 1, 11, or 20, and dependent claims 6 and 16 add only the Fourier transformation methodology without a termination condition. A person of ordinary skill in the art cannot determine with reasonable certainty when the limitation "the set of second training data is complete" is satisfied. See MPEP § 2173.05(b).
Dependent claims 2, 3, 5, 6, 7, 9, 10 (depending from claim 1), claims 12, 13, 14, 15, 16, 17, 18, 19 (depending from claim 11), and claim 20 inherit these indefiniteness defects from their respective independent claims.
For purposes of examination, "significant" is interpreted under BRI to encompass any element of the second training data whose inclusion or exclusion has a meaningful effect on the prediction performance of the first time-series machine-learning model, using any criterion. "Complete" is interpreted under BRI to mean that no additional beneficial time lag values need to be added to the second training data to further improve prediction performance, using any methodology.
Claim 11 recites in its preamble "a time-series data forecasting system" and the body continues with "the system comprising." However, later in the claim body, the limitation states "wherein the computer system is capable of performing a method comprising." There is insufficient antecedent basis for the term "the computer system" in claim 11. The claim introduces "a time-series data forecasting system" and uses "the system" to refer to that entity, but never introduces "a computer system" from which "the computer system" could derive antecedent basis. A person of ordinary skill in the art cannot determine with reasonable certainty whether "the computer system" is the same as "the time-series data forecasting system" (which would make the limitation redundant and confusing) or whether it refers to a different, unintroduced entity. See MPEP § 2173.05(e).
For purposes of examination, "the computer system" in claim 11 is interpreted under BRI to refer to the time-series data forecasting system as defined by the structural elements of claim 11 (processors, memories, storage media, and program instructions), consistent with the specification at ¶[0099] (Fig. 8). This interpretation requires equating two terms not identical on the claim's face.
Claims 4 and 14 each recite that the regularized machine-learning model is "selected from the group consisting of a Least Absolute Shrinkage and Selection Operator regression algorithm, a Ridge regression algorithm an Elastic Net regression algorithm, and a tree-based machine-learning model." The phrase "a Ridge regression algorithm an Elastic Net regression algorithm" is missing a comma (or conjunctive "and") between its two apparent components. As a result, the phrase can be read as (a) two separate Markush members — a Ridge regression algorithm and an Elastic Net regression algorithm — with an omitted separator, or (b) a single, unitary Markush member defined as a hybrid "Ridge regression algorithm [and] Elastic Net regression algorithm." Because a Markush group must unambiguously identify its members (MPEP § 2173.05(h)), and because the number and identity of the Markush members cannot be determined with reasonable certainty from the claim language alone, claims 4 and 14 are indefinite. Although the specification at ¶[0047] lists Ridge and Elastic Net as separate algorithms, claim language controls and the ambiguity cannot be resolved by importing limitations from the specification. See MPEP § 2173.05(h).
For purposes of examination, the Markush group is interpreted under BRI to contain four separate members: (1) a Least Absolute Shrinkage and Selection Operator regression algorithm, (2) a Ridge regression algorithm, (3) an Elastic Net regression algorithm, and (4) a tree-based machine-learning model, consistent with the specification at ¶[0047] which describes each as a distinct algorithm.
Claim 8 recites: "wherein the peak value is a plurality of a predefined number of peak frequency values of the Fourier transformation." First, the clause contains an internal logical inconsistency: parent claim 7 recites "a peak value" in the singular, and claim 8 states that "the peak value" (singular, inheriting the antecedent from claim 7) "is a plurality" (implicitly more than one). A singular entity cannot be "a plurality" without self-contradiction. A person of ordinary skill in the art cannot determine with reasonable certainty whether claim 8 is (a) redefining the single peak value of claim 7 as a set of multiple peak values, or (b) introducing a distinct, new configuration in which the peak value concept of claim 7 is replaced by a set of peaks. The claim is internally contradictory in this respect. See MPEP § 2173.
Second, the qualifier "predefined" in "a predefined number of peak frequency values" renders the limitation uncertain. The claim does not specify by whom the number is predefined, at what point in the process it is predefined, or what constraints apply to permissible values. The term "predefined" is a relative qualifier without an objective standard. See MPEP § 2173.05(b).
For purposes of examination, claim 8 is interpreted under BRI (drawing on specification at ¶[0052]) to mean that the single peak value of claim 7 is replaced by a plurality of peak frequency values of the Fourier transformation, with the number of such peaks being a user-specified or algorithm-specified integer determined prior to execution of the Fourier transformation step. This interpretation resolves the singular/plural contradiction by reading the claim as extending claim 7 to a multi-peak embodiment.
Claims 9 and 18 each recite: "wherein the first performance indicator, a second performance indicator and a third performance indicator are selected from the group consisting of..."
First, "the first performance indicator" (without additional term "value") lacks precise antecedent basis. Claims 1 and 11 (from which claims 9 and 18 respectively depend) each introduce "a first performance indicator value" (emphasis added). Claims 9 and 18 use "the first performance indicator" without "value," creating a terminological mismatch that fails to clearly trace the limitation to its antecedent. A person of ordinary skill in the art cannot determine with reasonable certainty whether "the first performance indicator" is identical to "a first performance indicator value" or refers to a distinct element. See MPEP § 2173.05(e).
Second, claims 9 and 18 introduce "a third performance indicator" without any antecedent, functional grounding, or operational context in the respective claim chains. Claims 1 and 11 (from which claims 9 and 18 directly depend, without passing through claims 5 or 15) use only one performance indicator (the first). Claims 9 and 18 introduce "a second performance indicator" and "a third performance indicator" but no step in the claim chains of claims 9 or 18 employs a third performance indicator in any evaluation, comparison, or optimization step. The third performance indicator is introduced but left functionally undefined and disconnected, rendering its scope uncertain. See MPEP § 2173.
For purposes of examination, "the first performance indicator" in claims 9 and 18 is interpreted under BRI as referring to "a first performance indicator value" from claims 1 and 11 respectively. "A second performance indicator" and "a third performance indicator" are interpreted under BRI as additional ML quality metrics of the same type, selectable from the listed group for use in any optimization step of the claimed method or system, consistent with the specification at ¶[0053].
Appropriate correction is required.
Conclusion
The most pertinent prior art(s) of record uncovered are: Autoregressive Process Modeling via the Lasso Procedure to Nardi et al. discloses: The Lasso is a popular model selection and estimation procedure for linear models that enjoys nice theoretical properties. In this paper, we study the Lasso estimator for fitting autoregressive time series models. We adopt a double asymptotic framework where the maximal lag may increase with the sample size. We derive theoretical results establishing various types of consistency. In particular, we derive conditions under which the Lasso estimator for the autoregressive coefficients is model selection consistent, estimation consistent and prediction consistent. Simulation study results are reported (Abstract); and Enhancing Peak Network Traffic Prediction via Time-Series Decomposition to Stewart et al. discloses: For network administration and maintenance, it is critical to anticipate when networks will receive peak volumes of traffic so that adequate resources can be allocated to service requests made to servers. In the event that sufficient resources are not allocated to servers, they can become prone to failure and security breaches. On the contrary, we would waste a lot of resources if we always allocate the maximum amount of resources. Therefore, anticipating peak volumes in network traffic becomes an important problem. However, popular forecasting models such as Autoregressive Integrated Moving Average (ARIMA) forecast time-series data generally, thus lack in predicting peak volumes in these time-series. More than often, a time-series is a combination of different features, which may include but are not limited to 1) Trend, the general movement of the traffic volume, 2) Seasonality, the patterns repeated over some time periods (e.g. daily and monthly), and 3) Noise, the random changes in the data. Considering that the fluctuation of seasonality can be harmful for trend and peak prediction, we propose to extract seasonalities to facilitate the peak volume predictions in the time domain. The experiments on both synthetic and real network traffic data demonstrate the effectiveness of the proposed method (Abstract).
Additional prior art has been made of record on the Notice of References Cited (Form PTO-892) and not relied upon, is considered pertinent to applicant's disclosure.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to ALAN CHEN whose telephone number is (571) 272-4143. The examiner can normally be reached M-F 10-7.
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, Kamran Afshar can be reached at (571) 272-7796. 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.
/ALAN CHEN/Primary Examiner, Art Unit 2125
Prosecution Insights