Microsoft Technology Licensing, LLCDownload PDFPatent Trials and Appeals BoardMar 29, 20222021004789 (P.T.A.B. Mar. 29, 2022) Copy Citation 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. 15/603,278 05/23/2017 Yifeng LIU 400418-US-NP 6234 69316 7590 03/29/2022 MICROSOFT CORPORATION ONE MICROSOFT WAY REDMOND, WA 98052 EXAMINER HSU, RYAN ART UNIT PAPER NUMBER 3715 NOTIFICATION DATE DELIVERY MODE 03/29/2022 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): chriochs@microsoft.com ljohnson@microsoft.com usdocket@microsoft.com PTOL-90A (Rev. 04/07) UNITED STATES PATENT AND TRADEMARK OFFICE BEFORE THE PATENT TRIAL AND APPEAL BOARD Ex parte YIFENG LIU, CIERRA D. MCDONALD, TIMOTHY ARTHUR GILL, and MARK ALLEN FINDLAY Appeal 2021-004789 Application 15/603,278 Technology Center 3700 Before DANIEL SONG, BRETT C. MARTIN, and LEE L. STEPINA, Administrative Patent Judges. STEPINA, Administrative Patent Judge. DECISION ON APPEAL STATEMENT OF THE CASE Pursuant to 35 U.S.C. § 134(a), Appellant1 appeals from the Examiner’s decision to reject claims 1-20. We have jurisdiction under 35 U.S.C. § 6(b). We AFFIRM IN PART. 1 We use the word “Appellant” to refer to “applicant” as defined in 37 C.F.R. § 1.42. Appellant identifies Microsoft Technology Licensing, LLC as the real party in interest. Appeal Br. 3. Appeal 2021-004789 Application 15/603,278 2 CLAIMED SUBJECT MATTER Appellant’s Application is titled “DETECTING CHEATING IN GAMES WITH MACHINE LEARNING.” Spec. 1. The Specification describes two examples (models) of machine learning, specifically, “principal components analysis” and “support vector machines.”2 Id. ¶ 21. As set forth in the Appeal Brief, “Appellant’s inventive contribution is the use of such models to detect game-level cheating via platform-level observations, which is implemented by training such models using the disclosed labeled training data and then classifying behaviors using the disclosed instance records.” Appeal Br. 32. Claim 1, reproduced below, is illustrative of the claimed subject matter. 1. A computing system comprising: a logic subsystem including a logic device; and a storage subsystem comprising a storage device, the storage subsystem comprising instructions executable by the logic subsystem to implement an online game platform hosting a plurality of games, each game of the plurality of games configured to send notifications to the game platform; implement a cheating detection module configured to, for a game available on the online game platform, classify user instance records for the game as normal or outlying via a trained machine-learning classification function, each user instance record comprising information provided by the game to the online game platform regarding achievements earned by a user of the game, and for each achievement a total game time it 2 In some instances, Appellant’s original disclosure uses the words “principle” and “principal” interchangeably. See, e.g., Spec. ¶¶ 21, 32, 35, 52, 53, Fig. 2. Appellant filed an Amendment to the Specification on September 24, 2020, that changes certain instances of “principle” to “principal.” Our citations refer to the original disclosure. Appeal 2021-004789 Application 15/603,278 3 took the user to earn the achievement, wherein user instance records classified as outlying represent possible cheating behavior, and for the game available on the online game platform, receive for each player of a plurality of players information from the game regarding notifications related to user progress in the game, the notifications provided by the game to the online game platform for each user of a plurality of users; for each user of the plurality of users, store an instance record for the user, the instance record for the user comprising information regarding achievements earned by the user in the game and a total game play time it took the user to earn each achievement earned by the user in the game, input the instance record for the user into the trained machine-learning classifying function to classify the instance record for the user as normal or as outlying, based at least in part on the instance record for the user being classified as outlying, then take an action based upon classification as outlying, and based at least in part on the instance record for the user not being classified as outlying, then not take the action. Appeal Br. 42-49 (Claims App.). Appeal 2021-004789 Application 15/603,278 4 EVIDENCE The evidence relied upon by Appellant is: Name Reference Date Appendix A EECE 580B - Support Vector Machines, Course Syllabus Spring 2010 Appendix B 12.3 Support Vector Machines and Kernels, pp. 423-536 Appendix C Cory Banks, GOG.com announces Galaxy, DRM- free online gaming platform, https://www.pcgamer.com June 5, 2014 Exhibit A Lab 15 - Support Vector Machines in Python Nov. 29, 2016 Exhibit B Ferhat Özgür Çatak, Mehmet Erdal Balaban, A MapReduce-based distributed SVM algorithm for binary classification, Turkish Journal of Electrical Engineering & Computer Sciences Dec. 11, 2013 REJECTIONS I. Claims 1-20 are rejected under 35 U.S.C. § 112(a) as failing to comply with the written description requirement. II. Claims 1-20 are rejected under 35 U.S.C. § 112(b) as indefinite. Appeal 2021-004789 Application 15/603,278 5 OPINION Invocation of 35 U.S.C. § 112(f) Independent Claim 1 The Examiner interprets the language “a cheating detection module” in claim 1 as reciting a generic placeholder that is coupled to functional language without sufficient structure to perform the recited function. Non- Final Act. 5-6. Therefore, the Examiner states, under 35 U.S.C. § 112(f), this language is “being interpreted to cover the corresponding structure described in the specification as performing the claimed function, and equivalents thereof.” Id. at 5. Appellant argues that “claim 1 is not properly interpreted under § 112(f), as the claim does not recite ‘means.’” Appeal Br. 18. Further, Appellant contends “claim 1 defines sufficient structure, both in the form of a specified set of algorithms to perform classification and also recited interactions between the cheating detection module and other components, to completely perform the function recited in the claim.” Id. According to Appellant, “[claim 1] recites that this classification is performed by structure, namely, a trained machine-learning classification function.” Id. at 19. According to Appellant, this recitation “excludes algorithms other than trained machine-learning based algorithms, such as hardcoded algorithms, and thus defines a species or sub-genus of structures capable of performing the classifying function that excludes an entire class of methods (non-trained functions).” Id. In response, the Examiner states, “the claim element ‘cheating detection module’ recited in claims 1-20 uses a generic placeholder (e.g., a nonce term or a non-structural term having no specific structural meaning) Appeal 2021-004789 Application 15/603,278 6 for performing the claimed function.” Ans. 3. According to the Examiner, “the recitation of the ‘cheating detection module’ is similar to the list of non- structural generic placeholders that may invoke 35 USC 112(f) that has been found to rebut the presumption that 112(f) does not apply.” Id. (citing MPEP 2181(I)(A)). As for the recitation of a trained machine-learning classification function in claim 1, the Examiner states, “[a]lthough machine learning techniques are algorithmic, without further disclosure as to how these general learning techniques models are transformed into a special purpose computer to perform the claimed function amount to pure functional claiming.” Id. at 5 (citing Aristocrat Tech. v. Intern. Game, 521 F.3d 1328, 1333 (Fed. Cir. 2008). The Examiner has the better position regarding interpretation of claim 1. “‘Module’ is a well-known nonce word that can operate as a substitute for ‘means’ in the context of § 112, para. 6 . . . ‘module’ is simply a generic description for software or hardware that performs a specified function.” Williamson v. Citrix Online, LLC, 792 F.3d 1339, 1350 (Fed. Cir. 2015). Appellant does not apprise us of any reason the words “cheating detection” before the word “module” connotes any particular structure distinguishing the combination of these words from a generic placeholder. The language in claim 1 following the “cheating detection module,” specifically, the phrase “configured to, for a game available on the online platform, classify user instance records for the game as normal or outlying via a trained machine-learning classification function,” is functional language, albeit limited to a particular context, i.e., for a game on an online platform and using a machine-learning function. In other words, this language specifies the function that the cheating detection module is Appeal 2021-004789 Application 15/603,278 7 configured to perform, namely, “classify[ing] user instance records for the game as normal or outlying via a trained machine-learning classification function.” As discussed in more detail below, the use of a trained machine- learning classification function, itself, requires an inquiry as to what structure is described in the Specification. The transition phrase “configured to” is recognized as a linking term that may invoke an interpretation under 35 U.S.C. § 112(f). “According to Rain, replacing ‘for’ with ‘configured to’ removed the means-plus-function language[, b]ut the purely functional claim language reciting what the ‘user identification module’ is configured to do provides no structure.” Rain Computing, Inc. v. Samsung Electronics America, Inc., 989 F.3d 1002, 1006 (2021) (citing MTD Prods. Inc. v. Iancu, 933 F.3d 1336, 1343 (Fed. Cir. 2019)). The remainder of the “cheating detection module” paragraph in claim 1 merely describes the information being used by the cheating detection module and such limitations upon the type of information used do not define the recited “module” in terms of structure. No other terms in claim 1, much less structural ones, modify the recited cheating detection module. Accordingly, we agree with the Examiner that this paragraph invokes an interpretation under 35 U.S.C. § 112(f) because it recites a nonce term followed by functional language without sufficient structure for performing the recited function. Rejection II (Indefiniteness) We realize that the scope of the claims must be addressed before addressing whether the Specification adequately supports them. “Any analysis [of a rejection for both indefiniteness and written description] should begin with the determination of whether the claims satisfy the Appeal 2021-004789 Application 15/603,278 8 requirements of the second paragraph [now 35 U.S.C. § 112(b)].” In re Moore, 439 F.2d 1232, 1235 (C.P.P.A. 1971). As the language at issue in claim 1 is properly interpreted as invoking 35 U.S.C. § 112(f), the rejections of claim 1 for indefiniteness and for failing to comply with the written description requirement presented in the Non-Final Office Action involve substantially the same issue, whether the means-plus-function limitation is provided with corresponding structure in the Specification and Figures. See Non-Final Act. 6-9. A rejection for indefiniteness “cannot stand where there is adequate description in the specification to satisfy [35 U.S.C. § 112(a)] regarding means-plus-function recitations in the claims that are not, per se, challenged for being unclear.” In re Noll, 545 F.2d 141, 149 (C.C.P.A. 1976). Accordingly, in addressing Rejection II (indefiniteness), we also address portions of the Examiner’s findings and Appellant’s arguments and citations to the record in relation to Rejection I (written description). The Examiner determines that claim 1 is indefinite because the “cheating detection module” and “trained machine-learning classification function” in claim 1 invoke 35 U.S.C. 112(f), but “the written description fails to disclose the corresponding structure, material, or acts for performing the entire claimed function and to clearly link the structure, material, or acts to the function.” Non-Final Act. 8. According to the Examiner, “the Specification fails to disclose the necessary steps and/or algorithm [as] to how the inventor achieved the functional limitation of ‘classify[ing] user instance records for the game as normal or outlying.’” Id. at 8. Therefore, the Examiner finds, “the Specification fails to clearly link the structure, material, or acts to the claimed function.” Id. at 9. Appeal 2021-004789 Application 15/603,278 9 In finding that the Specification fails to comply with the written description requirement, the Examiner takes issue with the following language in claim 1: a cheating detection module configured to, for a game available on the online game platform, classify user instance records for the game as normal or outlying via a trained machine-learning classification function, each user instance record comprising information provided by the game to the online game platform regarding achievements earned by a user of the game, and for each achievement a total game time it took the user to earn the achievement, wherein user instance records classified as outlying represent possible cheating behavior. Non-Final Act. 6-8. Specifically, the Examiner finds: the Specification fails to disclose in sufficient detail any algorithm for how the classifying function performed the claimed function the cheating detection module configured to detect possible instances of game-level cheating that can affect items awarded by the game platform and how the cheating detection module classifies the user progress in the game as normal or as outlying based upon the notifications is achieved. Id. at 7. The Examiner further finds that the Specification discloses the step of determining the total game time it took a user to earn an achievement, “but does not disclose how the cheating detection module performs analysis in relation to each achievement[, and how the] ‘total game time it took the user to earn the achievement’ [is used] to classify the records as ‘outlying’ to represent possible cheating behavior.” Id. Thus, the Examiner concludes the Specification fails to describe the corresponding steps and/or an algorithm to classify the user instance records as normal or outlying. Id. at 8. Appeal 2021-004789 Application 15/603,278 10 Appellant contends that the Specification3 “discloses examples of suitable machine learning functions that can be trained using labeled instance records to classify user instance records as normal or outlying. These example functions are support vector machines and principal components analysis.” Appeal Br. 20 (emphasis added). According to Appellant, “[b]oth of these techniques are algorithmic, and were well-known in the art prior to Appellant’s filing date, as shown by documentary evidence submitted by Appellant.” Id. In support of this assertion, Appellant refers to Appendix B and Exhibit B. See id. Appellant argues that the Application discloses “flow charts and text . . . that set forth how to classify user instance records as normal or outlying by first training a machine learning function with labeled instance records having known outcomes, and then using the trained machine learning function to classify possible behaviors.” Id. at 21. According to Appellant, “[Figure 2] shows a detailed flow chart setting forth how this classification is performed.” Id. Appellant then describes Figure 2, and with respect to step 228, states: a classification function is applied to classify the progress as normal or outlying, as indicated at 228. Examples include principal components analysis models 230 and support vector machine models 232. The output of the application of the function to the instance record is a decision whether the record is outlying, at 236. This decision can be in the form of a flag, and can include a confidence score. 3 In contesting both the indefiniteness rejection and written description rejection of claim 1, Appellant argues that the Specification discloses algorithms that a person of ordinary skill in the art would recognize as structure for performing the function of the claimed cheating detection module. See Appeal Br. 20-32; Reply Br. 2-12. Appeal 2021-004789 Application 15/603,278 11 Id. at 23 (citing Spec. ¶ 21). Appellant refers to paragraphs 22, 29-33 of the Specification as describing Figure 2. Id. Appellant also relies on Figure 3 and paragraphs 34-37 as describing a training process. See id. at 23-24. On page 31 of the Appeal Brief, Appellant attempts to map certain portions of claim 1 to paragraphs 24-25, 29, 31-37 of the Specification and Figures 2 and 3, stating “Appellant’s [S]pecification describes a finite sequence of steps for performing the task of classifying user instance records as normal or outlying using a trained machine learning function.” Appeal Br. 31. Appellant also refers to evidence submitted during prosecution, stating, “Appellant submitted documentary evidence showing the well- known status of support vector machines and principal components analysis. Mar. 16, 2020 Office Action Response, Appendices A-B.” Appeal Br. 32. In response, the Examiner acknowledges that the Specification discloses a (i) support vector machine model and (ii) principal components analysis, but finds “these are merely techniques to train a classifying function and not the algorithm as to how the inventor intended to classify user instance records as normal and outlying with respect to detecting cheating behavior in the cheating detection module.” Ans. 4-5. According to the Examiner, “[a]lthough machine learning techniques are algorithmic, without further disclosure as to how these general learning techniques models are transformed into a special purpose computer to perform the claimed function amount to pure functional claiming.” Id. at 5 (citing Aristocrat Tech., 521 F.3d at 1333) (emphasis added). The Examiner reiterates, “the Specification discloses machine learning techniques to train a function[, but] it does not provide the particular algorithm to achieve the classifying function used to classify instance records as normal or outlying.” Appeal 2021-004789 Application 15/603,278 12 Id. at 6 (citing Spec. ¶¶ 32-33) (emphasis added). The Examiner determines that the Specification does not sufficiently define structure and boundaries for the means-plus-function limitation in claim 1 to one of ordinary skill in the art because it fails to provide sufficient detail as to the rules for the classifying function regarding how to use (i) principal components analysis and (ii) support vector machines to classify user instance records. Id. According to the Examiner, Appellant has merely disclosed a class of algorithms for performing the functional language in claim 1. Id. at 8-11. For the following reasons, we sustain the rejection of claim 1 as indefinite. As discussed above, the function performed by the cheating detection module of claim 1 is “[to] classify user instance records for the game as normal or outlying via a trained machine-learning classification function.” The recited computing system of claim 1 must be a special purpose computer because specialized functions are recited following the “cheating detection module” language. In such cases, legal precedent requires that the specification disclose an algorithm for performing the claimed function. Williamson v. Citrix Online, LLC, 792 F.3d 1339, 1352 (Fed. Cir. 2015). The Federal Circuit has defined an algorithm as “a step- by-step procedure for accomplishing a given result.” Triton Tech of Texas, LLC v. Nintendo of Am., Inc., 753 F.3d 1375, 1379 (Fed. Cir. 2014) (quoting Ergo Licensing LLC v. CareFusion 303, Inc., 673 F.3d 1361, 1365 (Fed. Cir. 2012)). Arguing that the Specification discloses an algorithm for the function performed by the cheating detection module, Appellant refers to the principal components analysis and support vector machines disclosed in the Specification as providing “examples of suitable machine learning functions Appeal 2021-004789 Application 15/603,278 13 that can be trained using labeled instance records to classify user instance records as normal or outlying.”4 Appeal Br. 20. The first occurrence of these two terms in Appellant’s Specification is in paragraph 21, which states: To detect cheating occurring in games 120, 122, and also in third party games 180, 184, the game platform 112 includes a cheating detection module 152. The cheating detection module 152 is configured to detect possible instances of cheating based upon patterns of information that games provide to the platform regarding game play. The cheating detection module 152 comprises classifying functions, illustrated as game 1 classifying function 166 and game n classifying function 168, trained via machine learning techniques, that receive inputs of data provided by the games to the platform regarding progress of players in the games. Examples of such information include scores obtained, achievements met, items awarded, and ranks achieved. The classifying functions may utilize any suitable machine learning methods to perform such classification, including but not limited to, principal components analysis and support vector machines. Spec. ¶ 21 (emphasis added). Thus, the principal components analysis and support vector machines are listed, but without any discussion of a step-by- 4 Claim 9 depends from claim 1 and recites, “the instructions executable to apply the trained machine-learning classifying function comprise instructions executable to apply one or more of a principal components analysis model and a support vector machine model.” Appeal Br. 44 (Claims App.) (emphasis added). Claim 15 depends from claim 11 and recites “applying the trained machine-learning classifying function comprises applying one or more of a principal components analysis model and a support vector machine model.” Id. at 46 (emphasis added). Thus, based on the doctrine of claim differentiation, the classification function recited in each of independent claims 1 and 11 would appear to allow for more than the use of a principal components analysis and/or support vector machines. Appeal 2021-004789 Application 15/603,278 14 step procedure for accomplishing the given result of classification. Furthermore, the Specification makes clear that the classifying function may use any suitable machine learning method, not just principal components analysis and support vector machines. The Specification next discusses these two concepts in paragraph 32, which states: Continuing, the cheating detection module 152 applies a classifying function, at 228, to classify an instance for a player and a game as normal or outlying, wherein the classifying function has previously been trained via machine learning. The classifying function may have been trained via any suitable machine learning technique. For example, the classifying function may represent a principle components analysis model 230 or a support vector machine model 232. Furthermore, any other data regarding user behaviors on the platform 234 may be used to classify an updated instance as normal or outlying. Spec. ¶ 32 (emphasis added). Accordingly, in this portion of the Specification, the principal components analysis and support vector machines are described as being examples of models that can be used to train the classifying function. This paragraph indicates that the classifying function may be trained by any suitable machine learning technique, not just principal components analysis and support vector machines. Paragraph 35 of the Specification provides disclosure similar to that in paragraph 21, indicating that principal components analysis and support vector machines are non-limiting examples of classifying models. See Spec. ¶ 35. Paragraphs 52 and 53 of the Specification also mention the concepts of principal components analysis and support vector machines, but these Appeal 2021-004789 Application 15/603,278 15 paragraphs add little or nothing to the disclosure inasmuch as they merely paraphrase the limitations recited in original claims 1-17. Appellant maps the classifying function to paragraphs 34-37, as a group. See Appeal Br. 31. These paragraphs describe Figure 3, which discloses that a classifying function is trained based on data labeled normal or outlying, but does not address the central issue in Examiner’s rejection, which is how the cheating detection module classifies the user progress in the game as normal or as outlying. In other words, these paragraphs and Figure 3 disclose that the classifying function learns to perform the function recited in claim 1, but they do not disclose how that function is performed. Figure 2 discloses various steps performed by the computing system of claim 1, as a whole, but, with respect to the function performed by the cheating detection module, merely indicates that classification of user progress as normal or outlying is performed by a “PRINCIPLE COMPONENTS ANALYSIS MODEL 230,” a “SUPPORT VECTOR MACHINE MODEL 232,” and by “[USING] OTHER DATA REGARDING USER BEHAVIORS ON PLATFORM 234.” In other words, the majority of Figure 2 is not pertinent to defining an algorithm or structure for the cheating detection module, and the portion of Figure 2 that is pertinent to this issue is merely redundant to what is already disclosed in paragraph 32 of the Specification. The disclosure of the use of principal components analysis and support vector machines, as tersely set forth in Appellant’s Specification and Figures, does not amount to the disclosure of an algorithm for the purposes of 35 U.S.C. § 112(f) and the cheating detection module recited in claim 1. Rather, as the Examiner correctly finds (see Ans. 8), principal components Appeal 2021-004789 Application 15/603,278 16 analysis and support vector machines are classes of algorithms, not step-by- step procedures for accomplishing a given result. See Triton Tech of Texas, LLC v. Nintendo of Am., Inc., 753 F.3d at 1379. “Disclosure of a class of algorithms ‘that places no limitations on how values are calculated, combined, or weighted is insufficient to make the bounds of the claims understandable.’” Id. (Finding that the disclosure of numerical integration in the specification was insufficient to provide an algorithm for the recited “integrator means”). The evidence submitted by Appellant undermines Appellant’s assertion that it is only well-known “detailed inner workings” of principal components analysis and support vector machines that the Specification omits, and that such details are unnecessary to support a finding that the Specification discloses an algorithm sufficient to satisfy the written description requirement.5 See Appeal Br. 24-28, 32; Reply Br. 4. Specifically, Appellant’s Appendix A is a syllabus for a 3-credit course at Binghamton University titled “Support Vector Machines.” See App. A, 1. The Course Description in the syllabus states: Support Vector Machines belong to the class of supervised machine learning techniques. Thanks to their high performance and good generalization abilities, SVMs are becoming more and more popular among researchers of many different fields ranging from computer science through financial modeling to bioinformatics. SVMs are applicable to many different tasks, including classification, regression, pattern recognition, data mining, and predictive control. 5 There is no dispute that principal components analysis and support vector machines are well-known. Appeal 2021-004789 Application 15/603,278 17 Id. at 2 (emphases added). Thus, support vector machines are applicable to “many different tasks,” only one of which is classification, and to many different fields (from computer science to financial modeling). Accordingly, the mere disclosure of support vector machines does not amount to the disclosure of a step-by-step process of classification generally, much less “classify[ing] user instance records for [a] game as normal or outlying” as recited in claim 1. We have considered all of Appellant’s assertions that the cheating detection module recited in claim 1 is linked to corresponding structure in Appellant’s original disclosure, but find them unavailing. Accordingly, we sustain the rejection of claim 1 under 35 U.S.C. § 112(b) as indefinite. Independent Claims 11 and 18 Independent claims 11 and 18 recite substantially similar functional language, without further structure in the claims, as discussed above regarding claim 1. See Appeal Br. 45, 47 (Claims App.). Appellant relies on the arguments discussed above regarding the rejection of claim 1 to contest the rejection of independent claims 11 and 18 as indefinite. See Appeal Br. 36-37, 39-40. Accordingly, for the same reasons, we sustain the rejection of claims 11 and 18 as indefinite. Dependent Claims 2-10, 12-17, 19, and 20 For each of dependent claims 2-10, 12-17, 19, and 20, Appellant combines the discussion of the rejections under 35 U.S.C. §§ 112(a) and (b). See Appeal Br. 32-39, 41-42. In this discussion, Appellant lists the limitations in the claims, characterizes the limitations of the respective claim, and states that the limitations “define additional structure” or “define Appeal 2021-004789 Application 15/603,278 18 further structure” and, therefore, the claim should not be interpreted under 35 U.S.C. § 112(f). See id. Although Appellant’s arguments are curt, the Examiner provides no response to these arguments in the Answer and does not specifically discuss the dependent claims in the Non-Final Office Action. See Ans. 3-12; Non- Final Act. 2-10. As we do not have the Examiner’s input regarding whether claims 2-10, 12-17, 19, and 20 invoke a means-plus-function interpretation, and it is not self-evident that the discussion above regarding claim 1 applies equally to claims 2-10, 12-17, 19, and 20, we do not sustain the rejection of these claims as indefinite.6 Rejection I (Written Description) Appellant relies on the arguments discussed above in contesting the rejection of claims 1-20 as failing to comply with the written description requirement with respect to the means-plus-function limitation recited in claims 1, 11, and 18. Appeal Br. 28-32. In view of the lack of support in Appellant’s Specification and Figures discussed above, we sustain the rejection of independent claims 1, 11, and 18 as failing to comply with the written description requirement. However, based on the lack of discussion by the Examiner in the Non-Final Office Action and Answer, we do not 6 Our decision to reverse the rejection of claims 2-10, 12-17, 19, and 20 should not be construed as an indication that we think these claims are allowable. Rather, this reversal is based solely on the lack of analysis in the Non-Final Office Action and Answer with respect to these claims and Appellant’s associated arguments. Appeal 2021-004789 Application 15/603,278 19 sustain the rejection of dependent claims 2-10, 12-17, 19, and 20 on the same basis.7 In the Answer, aside from basing the rejection for lack of written description on a failure to disclose structure corresponding to a means-plus- function limitation in claim 1, the Examiner also finds that the Specification fails to support the full scope of “the claims” (claims 1, 11, and 18).8 Ans. 9-10. Specifically, the Examiner states: even if this [the disclosure of principle components analysis model and support vector machine model] is sufficient to disclose a trained classifying function it does not teach one of ordinary skill in the art how the inventor intended to transform these general models into a trained machine learning classifying function to detect cheating for the claimed cheating detection module. These necessary functions are not adequately described to show that the inventor had possession to implement a cheating detection module to classify user instance records of the online game platform “comprising information provided by the game to the online game platform regarding achievements earned by a user of the game, and for each achievement of a total game time it took the user to earn the achievement, wherein user instance records classified as outlying represent possible cheating behavior” (see Claim 1). Stated differently, the Specification 7 Our decision to reverse the rejection of claims 2-10, 12-17, 19, and 20 should not be construed as an indication that we think these claims are allowable. Rather, this reversal is based solely on the lack of detail in the Non-Final Office Action and Answer with respect to these claims and Appellant’s associated arguments. 8 Based on the context of the Examiner’s discussion, we understand the Examiner’s reference to “the claims” to mean only claims 1, 11, and 18 (and not all of claims 1-20). See Ans. 9-10 (“The claims recite[] a functional limitation, as exemplified in claim 1, to implement a cheating detection module configured to ‘for a game available on the online game platform, classify user instance records for the game as normal or outlying via a trained machine-learning classification function.’”). Appeal 2021-004789 Application 15/603,278 20 does not support the scope of the expansive claim language because the mere general statement of a generic method to be achieved via machine learning techniques does not entitle the inventor to claim any and all means for achieving that objective. Ans. 10 (citing LizardTech, Inc. v. Earth Resource Mapping, Inc., 424 F.3d 1336, 1346 (2005)) (emphasis added). Appellant asserts that LizardTech is “factually nonanalogous to the present case.” Reply Br. 12. According to Appellant, In Lizardtech, claim 21 was directed to creating a seamless array of DWT coefficients generically. Id. at 1344. In contrast, claim 1 was directed to a specific method of creating seamless DWT coefficients, by “maintaining updated sums.” Id. The specification disclosed only this method of creating DWT coefficients. Id. The court noted that claim 21 was broader than claim 1 because claim 21 omitted the “maintaining updated sums” element. Id. The court further noted that interpreting claim 21 to read only on the “maintaining updated sums” would be impermissibly reading a limitation into claim 21 to make it redundant with claim 1. Id. As such, the court found the full breadth of claim 21 not supported by the written description. Id. at 1345. Thus, in Lizardtech, a claim that omitted an element found in a similar claim was determined to have inadequate specification support because rules of claim interpretation required the two claims to have different scopes. Here, no such issue is presented. As such, the present case and Lizardtech are not factually analogous. Reply Br. 12. Appellant’s argument does not apprise us of Examiner error. As discussed above, Appellant’s Specification indicates that “[t]he classifying functions may utilize any suitable machine learning methods to perform such classification, including but not limited to, principal components analysis and support vector machines.” Spec. ¶ 21 (emphasis added); see Appeal 2021-004789 Application 15/603,278 21 also id. ¶¶ 32, 35. Appellant accurately summarizes a portion of the court’s claim interpretation in LizardTech determining that claim 21 was broader than claim 1. However, the process by which the court determined that claim 21 in LizardTech was broader than claim 1 is not pertinent to the Examiner’s finding that the full scope of claim 1 in the present case is not supported by the Specification. In other words, the fact that part of the claim interpretation process in LizardTech involved a comparison of the scope of two claims does not mean that the Examiner’s rejection of claim 1 in the present case must include a similar comparison in order to properly rely on the legal principle in LizardTech that § 112 precludes an inventor from reciting a claim scope far greater than what a person of ordinary skill in the art would understand the inventor to have possessed. See LizardTech, 424 F.3d at 1344-1346. Accordingly, Appellant’s argument does not address the Examiner’s rejection, and we sustain the rejection of claims 1, 11, and 18 on this alternative ground as failing to comply with the written description requirement. CONCLUSION The Examiner’s rejections are affirmed in part. DECISION SUMMARY Claims Rejected 35 U.S.C. § Reference(s)/Basis Affirmed Reversed 1-20 112(a) Written Description 1, 11, 18 2-10, 12- 17, 19, 20 1-20 112(b) Indefiniteness 1, 11, 18 2-10, 12- 17, 19, 20 Appeal 2021-004789 Application 15/603,278 22 Claims Rejected 35 U.S.C. § Reference(s)/Basis Affirmed Reversed Overall Outcome 1, 11, 18 2-10, 12- 17, 19, 20 TIME PERIOD FOR RESPONSE No time period for taking any subsequent action in connection with this appeal may be extended under 37 C.F.R. § 1.136(a). AFFIRMED IN PART Copy with citationCopy as parenthetical citation