Numenta, Inc.

Patent Trials and Appeals Board

Feb 24, 2022

2020005625 (P.T.A.B. Feb. 24, 2022)

UNITED STATES PATENT AND TRADEMARK OFFICE UNITED STATES DEPARTMENT OF COMMERCE United States Patent and Trademark Office Address: COMMISSIONER FOR PATENTS P.O. Box 1450 Alexandria, Virginia 22313-1450 www.uspto.gov APPLICATION NO. FILING DATE FIRST NAMED INVENTOR ATTORNEY DOCKET NO. CONFIRMATION NO. 14/494,324 09/23/2014 Jeffrey C. Hawkins II 25172-25807/US 7053 758 7590 02/24/2022 FENWICK & WEST LLP SILICON VALLEY CENTER 801 CALIFORNIA STREET MOUNTAIN VIEW, CA 94041 EXAMINER GONZALES, VINCENT ART UNIT PAPER NUMBER 2124 NOTIFICATION DATE DELIVERY MODE 02/24/2022 ELECTRONIC Please find below and/or attached an Office communication concerning this application or proceeding. The time period for reply, if any, is set in the attached communication. Notice of the Office communication was sent electronically on above-indicated "Notification Date" to the following e-mail address(es): PTOC@Fenwick.com PTOL-90A (Rev. 04/07) UNITED STATES PATENT AND TRADEMARK OFFICE ____________ BEFORE THE PATENT TRIAL AND APPEAL BOARD ____________ Ex parte JEFFREY C. HAWKINS II and SUBUTAI AHMAD ___________ Appeal 2020-005625 Application 14/494,324 Technology Center 2100 ____________ Before CARL W. WHITEHEAD JR., JEREMY J. CURCURI and ADAM J. PYONIN, Administrative Patent Judges. WHITEHEAD JR., Administrative Patent Judge. DECISION ON APPEAL1 Appellant2 is appealing the final rejection of claims 1, 3-13 and 15- 20 under 35 U.S.C. § 134(a). See Appeal Brief 6. Claims 1, 13 and 20 are independent. We have jurisdiction under 35 U.S.C. § 6(b). 1 Rather than reiterate Appellant’s arguments and the Examiner’s determinations, we refer to the Appeal Brief (filed December 30, 2019), the Reply Brief (filed May 11, 2020), the Final Action (mailed May 14, 2019) and the Answer (mailed April 20, 2020), for the respective details. 2 Appellant refers to “applicant” as defined in 37 C.F.R. § 1.42(a). (“The word ‘applicant’ when used in this title refers to the inventor or all of the joint inventors, or to the person applying for a patent as provided in §§ 1.43, 1.45, or 1.46.”). Appellant identifies Numenta, Inc. as the real party in interest. Appeal Brief 3. Appeal 2020-005625 Application 14/494,324 2 We affirm. Introduction According to Appellant, “The disclosure relates to modeling and presenting information regarding whether a target system is in an anomalous state based on the accuracy of predictions made by a predictive model.” Specification ¶ 2. Representative Claim3 (disputed limitations emphasized) 1. A method of detecting anomaly in a target system, comprising: receiving, by a processor, input data associated with the target system, the input data changing over time; based on the received input data of a first time, generating, by the processor, a prediction for a second time subsequent to the first time by executing one or more temporally predictive computer models on the received input data of the first time; generating, by the processor, a current accuracy score by comparing the prediction with an actual value corresponding to the prediction, the current accuracy score representing accuracy of the prediction made by the predictive computer models; and determining, by the processor, an anomaly score representing likelihood that the target system is in an anomalous state based on the current or one or more recent accuracy scores by referencing an anomaly model representing an anticipated range, or 3 Appellant does not argue independent claims 1, 13 and 20 individually. See Appeal Brief 9 (“Independent claims 13 and 20 recite similar features as claim 1. Therefore, claims 13 and 20 are also patentably distinguishable from the combination of the cited references.”). Accordingly, we select independent claim 1 as representative. See 37 C.F.R. § 41.37(c)(1)(iv) (“When multiple claims subject to the same ground of rejection are argued as a group or subgroup by appellant, the Board may select a single claim from the group or subgroup and may decide the appeal as to the ground of rejection with respect to the group or subgroup on the basis of the selected claim alone.”). Appeal 2020-005625 Application 14/494,324 3 distribution, of accuracy scores made by the predictive computer models, wherein the anomaly model is a distribution model that classifies the accuracy score into discrete ranges; processing the anomaly score for presenting information on the anomalous state of the target system to the user by comparing the anomaly score to a threshold value; retaining the one or more temporally predictive computer models after processing the anomaly score that indicates the target system is in the anomalous state; and generating, by the processor, a prediction for a third time subsequent to the second time by executing the retained one or more temporally predictive computer models. References Name4 Reference Date McGee US 2003/0088542 A1 May 8, 2003 Suyama US 2007/0299798 A1 December 27, 2007 Golic US 2010/0284282 A1 November 11, 2010 Hawkins US 2011/0225108 A1 September 15, 2011 Li No Pane, No Gain: Efficient Evaluation of Sliding-Window Aggregates over Data Streams March 2005 Perdisci Using an Ensemble of One-Class SVM Classifiers to Harden Payload-based Anomaly Detection Systems 2006 Chandola Anomaly Detection: A Survey July 2009 Rejections on Appeal5 4 All reference citations are to the first named inventor only. 5 The 35 U.S.C. § 101 rejection of claims 1, 3-13, and 15-20 is withdrawn. See Answer 4. Appeal 2020-005625 Application 14/494,324 4 Claims 1, 3, 12, 13, 15 and 20 stand rejected under 35 U.S.C. § 103 as being unpatentable over Suyama and Chandola. Final Action 6-10. Claims 5-7 and 17-19 stand rejected under 35 U.S.C. § 103 as being unpatentable over Suyama, Chandola and Golic. Final Action 10-11. Claims 8 and 9 stand rejected under 35 U.S.C. § 103 as being unpatentable over Suyama, Chandola and Li. Final Action 11. Claim 10 stands rejected under 35 U.S.C. § 103 as being unpatentable over Suyama, Chandola and Hawkins. Final Action 12. Claim 11 stands rejected under 35 U.S.C. § 103 as being unpatentable over Suyama, Chandola and Perdisci. Final Action 12-13. ANALYSIS In regard to claim 1, Appellant describes the claim as requiring that “[t]he predictive computer models are retained after the information is processed such that the user can first investigate the source of the anomaly and take remedial actions if needed, rather than simply accommodating the anomaly.” Appeal Brief 6 (citing Specification ¶¶ 62, 70-73). Note, none of the recited paragraphs explicitly discloses retaining the predictive computer models; paragraph 70 discloses a non-limiting example6 (emphasis added), “To facilitate the users to understand the relative significance of the anomaly parameters and trend, the anomaly parameters may be aggregated and presented to users in a graphical user interface such as a bar chart.” See Appeal Brief 6-7 (citing Specification ¶ 71 (emphasis added), Figures 4, 5). 6 See Phillips v. AWH Corp., 415 F.3d 1303, 1323 (Fed. Cir. 2005) (“[A]lthough the specification often describes very specific embodiments of the invention, we have repeatedly warned against confining the claims to those embodiments.”). Appeal 2020-005625 Application 14/494,324 5 Appellant argues that both Chandola and Suyama fail to disclose the claim limitation, “generating, by the processor, a prediction for a third time subsequent to the second time by executing the retained one or more temporally predictive computer models” as recited in claim 1. Appeal Brief 7-8. Appellant indicates support for the limitation is found in Figures 4, 5 and paragraphs 70-74 of the Specification. See Appeal Brief 3-4. The cited figures as well as the cited paragraphs do not explicitly disclose generating a prediction by executing retained temporally predictive models in the same manner recited in claim 1. Specifically, the cited portions of Appellant’s disclosure do not explicitly disclose second and third predictions as recited in claim 1. However, generally, paragraph 73 discloses “By enabling the user to view aggregated versions of the anomaly parameters over a desired time period, the user can quickly perceive the needed information” and “[t]he bar charts or other similar graphical user interfaces enable[ ] the users . . . to easily perceive trends and occurrences of anomalous states in the target system 102.” We note, the cited claim language must be given its broadest reasonable interpretation consistent with Appellant’s disclosure, as explained in Morris: [T]he PTO applies to the verbiage of the proposed claims the broadest reasonable meaning of the words in their ordinary usage as they would be understood by one of ordinary skill in the art, taking into account whatever enlightenment by way of definitions or otherwise that may be afforded by the written description contained in the applicant's specification. In re Morris, 127 F.3d 1048, 1054 (Fed. Cir. 1997); see also In re Zletz, 893 F.2d 319, 321 (Fed. Cir. 1989) (stating that “claims must be interpreted as broadly as their terms reasonably allow.”). Our reviewing court further states, “the proper ... construction is not just the broadest construction, but rather the Appeal 2020-005625 Application 14/494,324 6 broadest reasonable construction in light of the specification.” In re Man Mach. Interface Techs. LLC, 822 F.3d 1282, 1287 (Fed. Cir. 2016). Appellant contends, “both cited portions of Chandola fail to disclose generating separate predictions for incoming data using temporally predictive computer models let alone processing an anomaly score based on these predictions, and retaining the temporally predictive computer models to generate a prediction for a third time subsequent to the second time.” Appeal Brief 7 (citing Chandola Sections 3.1, 3.2.1). Appellant argues, “At most, the cited portions of Chandola appear to detect anomalies using the incoming data itself, rather than generating separate predictions that correspond to actual values of the incoming data to determine accuracy scores.” Appeal Brief 7. Appellant arguments are unpersuasive, because the Examiner relies upon Suyama and not Chandola to disclose the separate predictions as well as a current accuracy score. See Final Action 6. The Examiner relies on Chandola for disclosing, “determining, by the processor an anomaly score representing likelihood that the target system is in an anomalous state based on the current or one or more recent accuracy scores.” Final Action 7. The Examiner determines Chandola discloses, “that the [computer] model must be retained if it is to be used continuously in online analysis, e.g.[,] as described at p. 11, sec. 3.1 (discussing online analysis of streaming data) or p. 13, sec. 3.2.1 (discussing online detection of fraud as soon as the fraudulent transaction takes place)).” Final Action 8 (emphasis added). Appellant contends, “However, such an assertion is conclusory at best, and is a hindsight reading of the cited reference.” Appeal Brief 7. Appellant argues that, “The Office Action fails to explain in detail why a person of skill in the art considering Chandola would find it obvious to retain a temporally predictive Appeal 2020-005625 Application 14/494,324 7 model after processing the anomaly score that indicates the target system is in the anomalous state.” Appeal Brief 7. Here, Appellant’s argument is not persuasive of Examiner error because the Examiner cites to Chandola sections 3.1/3.2.1 and Appellant does not argue the merits of Chandola’s cited disclosure. See 37 C.F.R. § 41.37(c)(1)(iv) (2013) (“The arguments shall explain why the examiner erred as to each ground of rejection contested by appellant.” (Emphasis added.)). Furthermore, “[i]t is not the function of [the Board] to examine the claims in greater detail than argued by an appellant, looking for [patentable] distinctions over the prior art.” In re Baxter Travenol Labs., 952 F.2d 388, 391 (Fed. Cir. 1991). Further, the Examiner determines: a person of ordinary skill in the art would have found it obvious to use a threshold and/or histogram approach to detect anomalies (as taught by Chandola) in a time-series prediction system (such as Suyama) because either thresholds or histogram binning can identify values which may indicate that a system is behaving anomalously. Final Action 8. Here, the Examiner articulates reasoning to combine the references supported by rational underpinning to support the legal conclusion of obviousness. See KSR Int’l. v. Teleflex Inc., 550 U.S. 398, 418 (2007) (citing In re Kahn, 441 F.3d at 988). Appellant contends, “Suyama fails to disclose that the error is presented to the user as information on an anomalous state of the time series data, nor does it disclose retaining the time series model and generating a prediction at a subsequent time using the retained time series model.” Appeal Brief 8. The Examiner relies upon Chandola and not Suyama to disclose this feature of the claims; thus Appellant does not show Examiner error. See Final Action 8 (citing Chandola pages 10, 11, 19, 23, 31 and Figures 6, 9). Appellant argues, “In contrast, Suyama discloses that ‘[i]f the error is larger than the preset Appeal 2020-005625 Application 14/494,324 8 error ... the model series candidate creation unit 35 creates a plurality of new model series’” and therefore “even if Suyama can somehow be combined with Chandola, the combination of Chandola and Suyama is still improper because Suyama teaches away from the recited feature.” Appeal Brief 8 (citing Suyama ¶ 33). Appellant’s argument is not persuasive of Examiner error because Appellant’s bare assertion does not show persuasively that Suyama actually teaches away from the recited feature. As articulated by the Federal Circuit, “A reference may be said to teach away when a person of ordinary skill, upon reading the reference, would be discouraged from following the path set out in the reference, or would be led in a direction divergent from the path that was taken by the applicant.” In re Kahn, 441 F.3d 977, 990 (Fed. Cir. 2006) (citations and internal quotation marks omitted). Suyama discloses, “The prediction error calculation unit 34 calculates an error between the new unit variate time series data and a prediction value estimated from the model series stored in the model series memory unit 33 (S14).” Suyama ¶ 33, Figures 1, 2 (emphasis added). Suyama further discloses, “Then, it calculates an error between the value calculated according to the equation (1) and the unit variate time series data read out from the time series data memory unit 31.” Suyama ¶ 33, Figure 1 (emphasis added). The Examiner relies upon Chandola to disclose processing the anomaly score for presenting information on the anomalous state of the target system as recited in claim 1. See Final Action 8 (citing Chandola pages 10, 11, 19, 23, 31 and Figures 6, 9). Therefore, contrary to Appellant’s argument, Suyama does not teach away from the recited feature. See Kahn, 441 F.3d at 990. Appeal 2020-005625 Application 14/494,324 9 We sustain the Examiner’s obviousness rejection of claims 1, 3-8, 10-13 and 15-20, not argued separately. See Appeal Brief 9-10. Appellant states, “Claim 9 depends from claims 1 and 8, and recites the feature of ‘increasing or decreasing a time period represented by the aggregated accuracy score responsive to receiving another user input.’” Appeal Brief 10. Appellant further states: The recited limitation is advantageous, among other reasons, because it allows the information on the anomalous state of the target system to be presented to the user in different time periods, such that user can quickly perceive the needed information and easily perceive trends and occurrences of anomalous states in the target system. Appeal Brief 10 (citing Specification ¶ 73; Figures 4, 5). Paragraph 73 of the Specification is silent in regard to the “increasing or decreasing of a time period represented by the aggregated accuracy score” recited in claim 9. Further, the paragraph does not discuss the advantages of increasing or decreasing a time period that Appellant states in the Appeal Brief. See Appeal Brief 10. Appellant also states, “For example, responsive to user input, an aggregated anomaly score over a period represented by an individual bar can be expanded to reveal a detailed breakdown of multiple aggregated anomaly parameters over a shorter period, as illustrated, for example, in FIGS. 4-5 of the application-as-filed.” Appeal Brief 10 (citing Specification ¶ 72; Figures 4, 5). Paragraph 72 provides an example of expanding an aggregated anomaly score however the example does not define nor distinguish the “increasing or decreasing of a time period represented by the aggregated accuracy score” limitation recited in claim 9. See Phillips, 415 F.3d at 1323. The Examiner determines, “Li discloses its further limitation Appeal 2020-005625 Application 14/494,324 10 comprising increasing or decreasing a time period represented by the aggregated accuracy score responsive to receiving another user input (p. 39, first col.: user can define or modify assignment to variable RANGE).” Final Action 11. Appellant argues, “Li discloses a method of evaluating sliding- window aggregate queries, see Li, Abstract, and does not disclose increasing or decreasing a time period represented by an aggregated accuracy score responsive to receiving a user input to indicate an anomalous state of the target system.” Appeal Brief 10-11. Claim 9 requires the time period to either be increased or decreased. Li discloses, “Sliding window aggregate queries allow users to aggregate the stream at a user-specified granularity (RANGE) and interval (SLIDE), and thus provide the users a flexible way to monitor streaming data.” Li page 39 (emphasis added). Accordingly, Li modifies the combination of Suyama and Chandola by incorporating flexible ways to monitor streaming data over time in the same manner as recited in claim 9 as supported by the Specification. We sustain the Examiner’s obviousness rejection of claim 9. CONCLUSION Claims Rejected 35 U.S.C. § Reference(s)/Basis Affirmed Reversed 1, 3, 12, 13, 15, 20 103 Suyama, Chandola 1, 3, 12, 13, 15, 20 5-7, 17-19 103 Suyama, Chandola, Golic 5-7, 17-19 8, 9 103 Suyama, Chandola, Li 8, 9 10 103 Suyama, Chandola, Hawkins 10 Appeal 2020-005625 Application 14/494,324 11 11 103 Suyama, Chandola, Perdisci 11 No time period for taking any subsequent action in connection with this appeal may be extended under 37 C.F.R. § 1.136(a)(1). See 37 C.F.R. § 1.136(a)(1)(v). AFFIRMED