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 .
DETAILED ACTION
Response to Amendment
In the amendment dated 01/02/2026, the following occurred: Claims 15, 17 and 19 are amended; and claims 5 and 16 were previously cancelled.
Claims 1-4, 6-15 and 17-24 are pending and have been examined.
Priority
This application claims priority to U.S. Provisional Patent Application Nos. 63/407,602 filed 9/16/2022 and 63/404,531 filed 9/07/2022.
Claim Interpretation
The following is a quotation of 35 U.S.C. 112(f):
(f) Element in Claim for a Combination. – An element in a claim for a combination may be expressed as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof.
The claims in this application are given their broadest reasonable interpretation using the plain meaning of the claim language in light of the specification as it would be understood by one of ordinary skill in the art. The broadest reasonable interpretation of a claim element (also commonly referred to as a claim limitation) is limited by the description in the specification when 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is invoked.
As explained in MPEP § 2181, subsection I, claim limitations that meet the following three-prong test will be interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph:
(A) the claim limitation uses the term “means” or “step” or a term used as a substitute for “means” that is a generic placeholder (also called a nonce term or a non-structural term having no specific structural meaning) for performing the claimed function;
(B) the term “means” or “step” or the generic placeholder is modified by functional language, typically, but not always linked by the transition word “for” (e.g., “means for”) or another linking word or phrase, such as “configured to” or “so that”; and
(C) the term “means” or “step” or the generic placeholder is not modified by sufficient structure, material, or acts for performing the claimed function.
Use of the word “means” (or “step”) in a claim with functional language creates a rebuttable presumption that the claim limitation is to be treated in accordance with 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph. The presumption that the claim limitation is interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is rebutted when the claim limitation recites sufficient structure, material, or acts to entirely perform the recited function.
Absence of the word “means” (or “step”) in a claim creates a rebuttable presumption that the claim limitation is not to be treated in accordance with 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph. The presumption that the claim limitation is not interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is rebutted when the claim limitation recites function without reciting sufficient structure, material or acts to entirely perform the recited function.
Claim limitations in this application that use the word “means” (or “step”) are being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, except as otherwise indicated in an Office action. Conversely, claim limitations in this application that do not use the word “means” (or “step”) are not being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, except as otherwise indicated in an Office action.
This application includes one or more claim limitations that use the word “means” or “step” (or a generic placeholder) but are nonetheless not being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph because the claim limitation(s) recite(s) sufficient structure, materials, or acts to entirely perform the recited function (reciting “a hybrid machine learning model that includes… a deterministic solver”). Such claim limitation(s) is/are: “a deterministic solver configured to solve” in claims 1, 14 and 19 (emphasis added and claim 1 being representative).
Because this/these claim limitation(s) is/are not being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, it/they is/are not being interpreted to cover only the corresponding structure, material, or acts described in the specification as performing the claimed function, and equivalents thereof.
If applicant intends to have this/these limitation(s) interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, applicant may: (1) amend the claim limitation(s) to remove the structure, materials, or acts that performs the claimed function; or (2) present a sufficient showing that the claim limitation(s) does/do not recite sufficient structure, materials, or acts to perform the claimed function.
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 1-4, 6-15 and 17-24 are rejected under 35 U.S.C. §101 because the claimed invention is directed to a judicial exception (i.e., an abstract idea) without significantly more.
Claims 1, 14 and 19 are rejected under 35 U.S.C. §101 because the claimed invention is directed to an abstract idea without significantly more.
Eligibility Analysis Step 1 (YES): Claims 1, 14 and 19 fall into at least one of the statutory categories (i.e., process, machine or manufacture).
Eligibility Analysis Step 2A1 (YES): The claims recite an abstract idea. The identified abstract idea is as underlined (claim 1 being representative):
receiving, by a portable electronic device prior to a user engaging in a future activity, a user input corresponding to the future activity;
providing, at the portable electronic device, a hybrid machine learning model that includes a trained first neural network and a deterministic solver, wherein:
the trained first neural network has been trained based at least in part on prior physiological data associated with prior activities performed prior to the future activity; and
the deterministic solver for a physiological state equation that includes the trained first neural network as a function within the physiological state equation;
providing, by the portable electronic device responsive to receiving the user input, activity information for the future activity to the deterministic solver of the hybrid machine learning model;
generating, by executing the hybrid machine learning model at the portable electronic device and based on the provided activity information, a physiological prediction for the user with respect to the future activity by determining, with the deterministic solver, a deterministic solution to the physiological state equation that includes the trained first neural network as the function; and
providing, by the portable electronic device, an output corresponding to the physiological prediction.
The identified limitations, as drafted, is a process that, under its broadest reasonable interpretation (BRI), covers a mental process alone or along with a mathematical concept that includes mathematical relationships, mathematical formulas or equations, and mathematical calculations (but for the recitation of generic computer component language discussed in Step 2A2). That is, other than reciting the generic computer component language, nothing in the claims precludes the step(s) from practically being performed in human mind(s) with or without use of a physical aid (e.g., pen and paper, slide rule) (MPEP §§ 2106.04(a) and (a)(2)(III)) alone, or along with providing a trained first neural network and a deterministic solver and solving a physiological state equation that encompasses a mathematical concept. For example, the claims encompass a person thinking about receiving information corresponding to a future activity prior to a user engaging in the future activity, providing activity information for the future activity to a deterministic solver (e.g., on pen and paper) and generating a physiological prediction for the user with respect to the future activity (e.g., by performing hand calculations). If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation in the mind with or without a physical aid, then it falls within the “Mental Processes” grouping of abstract ideas.
The claims further recite the encompassed mathematical concept. For example, but for the generic computer component language, the claims recite (1) a trained neural network that has been trained outside the scope of the claim, i.e., a mathematical model or function; and (2) solving, with the deterministic solver (which does not connote structure), the physiological state equation in the manner described in the identified abstract idea, supra. When given its broadest reasonable interpretation in light of the disclosure, the encompassed mathematical concept represents the creation of mathematical interrelationships between data (see Spec. at para. 0033-0034, 0036). As such, the application of the trained neural network and the solving of the physiological state equation (that includes the trained first neural network as the function) represent mathematical concept(s) that is/are interpreted to be part of the identified abstract idea, supra. The Examiner notes that the mathematical concept need not be expressed in mathematical symbols but not merely limitations that are based on or involve a mathematical concept (MPEP § 2106.04(a)(2)(I)).
Accordingly, the claims recite multiple abstract ideas, which fall in different groupings. See MPEP § 2106.
The Examiner notes that the abstract idea could also be characterized as a certain method of organizing human activity (managing personal behavior or relationships or interactions between people) alone or along with a mathematical concept, reciting multiple abstract ideas falling into different abstract idea sub-groupings; however, this has been omitted for brevity as it is not required.
Eligibility Analysis Step 2A2 (NO):
The judicial exception, the above-identified abstract idea, is not integrated into a practical application. In particular, the claims recite the additional elements of a portable electronic device (claims 1, 14, 19) comprising a memory and one or more processors (claim 14) and a non-transitory machine-readable medium implemented by a processor (claim 19) that implement the identified abstract idea. The additional elements aforementioned are not described by the applicant and are recited at a high-level of generality (i.e., a generic computer or computer component performing a generic computer or computer component function that facilitates the identified abstract idea) such that these amount no more than mere instructions to apply the exception using a generic computer component (see Applicant’s disclosure at Fig. 1, and at para. 0025, 0087). See MPEP § 2106.04(d)(I). Accordingly, even in combination, these additional elements do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea. The claims are directed to an abstract idea.
The claims further recite the additional element of a hybrid machine learning model that implements the identified abstract idea. The additional element is not described by the Applicant, is recited at a high-level of generality and is merely invoked as a tool to perform an existing process (MPEP § 2106.05(f)(2), see case involving a commonplace business method or mathematical algorithm being applied on a general-purpose computer within the “Other examples”), such that this amounts no more than mere instructions to apply the abstract idea using a general-purpose computer (see Specification, e.g., at para. 0087). See MPEP § 2106.04(d)(I); and Alice Corp. Pty. Ltd. V. CLS Bank Int’l, 134 S. Ct. 2347, 1357 (2014). Accordingly, even in combination, these additional elements do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea. The claims are directed to an abstract idea.
Alternately, the claims further recite the presumed additional element of a trained neural network that implements the identified abstract idea (to the extent that the trained neural network requires a computer to be implemented). The additional element is not described by the Applicant, is recited at a high-level of generality and is merely invoked as a tool to perform an existing process (MPEP § 2106.05(f)(2), see case involving a commonplace business method or mathematical algorithm being applied on a general-purpose computer within the “Other examples”), such that this amounts no more than mere instructions to apply the abstract idea using a general-purpose computer (see Applicant’s disclosure at Fig. 3-4 and para. 0019, 0033-0035, 0045). See MPEP § 2106.04(d)(I); and Alice Corp. Pty. Ltd. V. CLS Bank Int’l, 134 S. Ct. 2347, 1357 (2014). Accordingly, even in combination, these additional elements do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea. The claims are directed to an abstract idea.
Eligibility Analysis Step 2B (NO):
The claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, the additional elements of a portable electronic device (claims 1, 14, 19) comprising a memory and one or more processors (claim 14) and a non-transitory machine-readable medium implemented by a processor (claim 19) to perform the method (represented by claim 1) amount no more than mere instructions to apply the exception using a generic computer or generic computer component. Mere instructions to apply an exception using generic computer(s) and/or generic computer component(s) cannot provide an inventive concept (“significantly more”). See MPEP § 2106.05(f).
The claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, the additional element of a hybrid machine learning model to perform the method amounts no more than mere instructions to “apply it” with the exception by invoking an algorithm merely as a tool to perform an existing process (i.e., only recites the algorithm as a tool to apply data to an algorithm and report the results), in this case to receive input data and generate output data. The use of a hybrid machine learning algorithm in its ordinary capacity to perform tasks in the identified abstract idea does not provide an inventive concept (“significantly more”). See MPEP § 2106.05(f). Accordingly, alone or in combination, the additional element does not provide significantly more.
Thus, the claims are not patent eligible.
Also in the alternative, the claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, the presumed additional element of a trained neural network to perform the method amounts no more than mere instructions to “apply it” with the exception by invoking an algorithm merely as a tool to perform an existing process (i.e., only recites the algorithm as a tool to apply data to an algorithm and report the results), in this case to receive input data and generate output data. The use of a trained neural network algorithm in its ordinary capacity to perform tasks in the identified abstract idea does not provide an inventive concept (“significantly more”). See MPEP § 2106.05(f). Accordingly, alone or in combination, the additional element does not provide significantly more. Thus, the claims are not patent eligible.
Dependent claims 2-4, 6-13, 15, 17-18 and 20-24, when analyzed as a whole, are similarly rejected under 35 U.S.C. §101 because the additional limitation(s) fail(s) to establish that the claim(s) is/are not directed to an abstract idea without significantly more. The claims, when considered alone or as an ordered combination, either (1) merely further define the abstract idea, (2) do not further limit the claim to a practical application, or (3) do not provide an inventive concept such that the claims are subject matter eligible.
Claim(s) 2, 15 and 20 merely further describe the abstract idea implemented by the hybrid machine learning model further comprising a second neural network trained as a user embedding model (e.g., generating data, providing data). See analysis, supra.
Assuming arguendo that the first (and second and third and fourth) neural networks are not mathematical concept(s) (as the first neural network is a function within the physiological state equation in representative claim 1), Claims 2, 11, 15, 17, and 20 further describe the additional elements of a trained second (and third and fourth) neural network and training the first neural network of the hybrid machine learning model that implements the identified abstract idea. The additional elements are not described by the Applicant, are recited at a high-level of generality and are merely invoked as tools to perform an existing process (MPEP § 2106.05(f)(2), see case involving a commonplace business method or mathematical algorithm being applied on a general-purpose computer within the “Other examples”), such that this amounts no more than mere instructions to apply the abstract idea using a general-purpose computer (see Specification, e.g., at para. 0087). See MPEP § 2106.04(d)(I); and Alice Corp. Pty. Ltd. V. CLS Bank Int’l, 134 S. Ct. 2347, 1357 (2014). Accordingly, even in combination, the additional elements do not (would not) integrate the abstract idea into a practical application because they do not (would not) impose any meaningful limits on practicing the abstract idea. The claims are directed to an abstract idea.
The claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, the (presumed) additional elements of a trained second (and third and fourth) neural network and training the first neural network of a hybrid machine learning model to perform the method amounts no more than mere instructions to “apply it” with the exception by invoking an algorithm merely as a tool to perform an existing process (i.e., only recites the algorithm as a tool to apply data to an algorithm and report the results), in this case to receive input data and generate output data. The use of a hybrid machine learning algorithm in its ordinary capacity to perform tasks in the identified abstract idea does not provide an inventive concept (“significantly more”). See MPEP § 2106.05(f). Accordingly, alone or in combination, the additional elements do not provide significantly more.
Claims 3-4, 6-10, 13 and 21-24 merely further describes the abstract idea (e.g., the historical activity information, the future activity, the physiological prediction, the environmental information, using a Gaussian process, the activity information, the user input, the output).
Claim 11 merely describes the additional element of training the first neural network (e.g., as a user-demand model, the trained first neural network being a mathematical model or function such as the user-demand model), invoking the training of the first neural network merely as a tool to perform an existing process, such that this amounts no more than mere instructions to apply the abstract idea using a general-purpose computer. See analysis, supra.
Claim 12 merely further describe the abstract idea implemented by the hybrid machine learning model. See analysis, supra.
Claim 17 merely further describes the abstract idea implemented by the hybrid machine learning model (e.g., the trained first neural network comprises a user-demand model, the trained third neural network comprises a fatigue model, ).
Claim 18 merely further describes the abstract idea (e.g., the physiological state equation further comprises the trained neural networks / the trained fatigue model and the trained weather-demand model as functions within the physiological state equation).
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 1-4, 6-15, 19-20 and 22-24 are rejected under 35 U.S.C. 103 as being unpatentable over Ni et al. (2019) (“Modeling Heart Rate and Activity Data for Personalized Fitness Recommendation”; “Ni” herein) in view of Morlock (US 2014/0303892 A1).
Re. Claim 1, Ni teaches a method comprising:
receiving, by […] prior to a user engaging in a future activity, a user input corresponding to the future activity (Abstract teaches wearable devices. Fig. 3 teach receiving user input.);
providing, at […], a hybrid machine learning model (see Fig. 3, Model structure) that includes a trained first neural network (a trained first LSTM) and a deterministic solver (FitRec for workout profile forecasting or FitRec-Attn for short term prediction), wherein:
the trained first neural network has been trained based at least in part on prior physiological data associated with prior activities performed prior to the future activity (Abstract and pg. 1344, Section 1. “Introduction” teach a Long Short-Term Memory (LSTM) based model FitRec that takes as input user attributes and multiple workout measurement sequences (prior physiological data associated with prior activities)… FitRec can then be applied to (1) Quantitative tasks, such as personalized sequential modeling to predict how workout measurements (e.g., heart rate) will change across a workout… beforehand (see Fig. 2, workout profile forecasting). Pg. 1345, Section 3.1 “Notation and Tasks” teaches the following (training): “Assume we are given N sequences X = (x1, … , xT) ∈ RN×T and one target sequence y = (y1, … ,yT ) ∈ RT , where xt =(x1t , x2t , … , xNt ) ∈ RN , yt ∈ R, and t ∈ [1,T]. Here X can be considered as a combination of various contextual sequences (e.g. distance, altitude) for the current workout and y is a target measurement sequence (e.g. heart rate, speed)”. Fig. 3 teaches historical user sequences input; and Table 2 teaches the sequence measurements, i.e., heart rate, timestamps, distance, speed. Pg. 1347, Section 3.3.1 “LSTM to process sequences” teaches an LSTM learning an update function (also training) … Subsequently, pg. 1348, Section 3.4 “Objective Function” teaches use of equation (12) (the first neural network has been trained) for both workout profile forecasting and short-term prediction.); and
the deterministic solver (2-layer stacked LSTM module) is configured to solve a physiological state equation (2-layer stacked LSTM model) that includes the trained first neural network (the trained first LSTM) as a function within the physiological state equation (Fig. 3, pg. 1344 at Section 1 “Introduction”, and pg. 1347 at Section 3.3.3 “FitRec” teach using a 2-layer stacked LSTM module/model for workout profile forecasting.);
providing, by […] responsive to receiving the user input, activity information for the future activity to the deterministic solver (2-layer stacked LSTM module) of the hybrid machine learning model (see Fig. 3 Model structure) (Fig. 3 and pg. 1344, Section 1. “Introduction” teach the LSTM-based model FitRec uses shared embedding layers while adapting a common predictive component: a two-layer stacked LSTM module (deterministic solver) for workout profile forecasting; takes as input user attributes and multiple workout measurement sequences (receiving the user input); and can then be applied to (1) Quantitative tasks, such as personalized sequential modeling (providing the activity information) to predict how workout measurements (e.g., heart rate) will change across a workout… beforehand (i.e., based on a map of the intended route) (see Fig. 2, workout profile forecasting).);
generating, by executing the hybrid machine learning model (Fig. 3 Model Structure) at […] and based on the provided activity information, a physiological prediction for the user with respect to the future activity by determining, with the deterministic solver, the physiological state equation that includes the trained first neural network as the function (see previous citations. Fig. 3 teaches the generated outputs (necessarily generated) are colored in blue.); and
providing, by […], an output corresponding to the physiological prediction (Fig. 3 teaches the generated outputs are colored in blue (necessarily provided). See additionally pg. 1346, Section 3.3 “Model Structure” and pg. 1347, Section 3.3.3 “FitRec”. See additionally Fig. 4 and associated text.)
Ni may not teach a portable electronic device.
Morlock teaches
receiving, by a portable electronic device (200) prior to a user engaging in a future activity, a user input corresponding to the future activity (Fig. 5, [0181] teach a user enters information into a personal portable training device 200 to setup a new trail activity prior to trail traversal. [0201], [0210] teaches information useful for trail routing comprises notable sensors or participant input including user profile information, e.g., physical parameters such as initial heart rate. See also Fig. 2, [0107]-[0109].);
providing, at the portable electronic device (200), a hybrid machine learning model that includes a trained first neural network… wherein: the trained first neural network has been trained based at least in part on prior physiological data associated with prior activities performed prior to the future activity (Figs. 1A-1B, [0080] teach developing a routing cost model (hybrid machine learning model) using machine learning techniques. Fig. 1A, [0047], [0086] teach constructing several neural networks based at least in part on heart rate sensor data during traversal 108 (prior physiological data associated with prior activities), training each network and determining weighting factors and biases for prediction of travel time per trail segment 110 (necessarily including a trained first network).);
providing, by the portable electronic device (see previous citations); and
generating, by executing the hybrid machine learning model at the portable electronic device (Figs. 1B, 5 teach using the model to predict travel time with an error estimate given the sensors suite used to characterize the participant 118 (necessarily generated)… and subsequent performance of route calculations 508 prior to starting trail traversal. Fig. 2-3, [0182] teach the device 200 includes a processor 202 connected to an input device 212 and a display screen 210.)
Therefore, it would have been prima facie obvious to one of ordinary skill in the art before the effective filing date to have modified the method of modeling heart rate and activity data for personalized fitness recommendation of Ni to utilize a personal portable training device for development and implementation of a route cost model and to use this information as part of a method and apparatus for creating cost data for use in generating a route across an electronic map as taught by Morlock, with the motivation of improving user health and safety during personal training exercises and sporting activities (see Morlock, e.g., at para. 0003-0005).
Re. Claim 2, Ni/Morlock teaches the method of claim 1, wherein the hybrid machine learning model further comprises a second neural network trained as a user embedding model, the method further comprising: generating, by the second neural network, a learned latent representation for the user, based at least on training data comprising historical activity information for the user (Morlock Fig. 1A, [0048] teaches training each of several neural networks using data which is known. Ni pg. 1346, Section 3.3 “Model Structure” teaches both models share the same basic structure consisting of a contextual embedding module and an attribute embedding layer (user embedding model)… These layers encode user attributes and historical information (training data) into an embedded representation (learned latent representation) that captures their latent individual attributes (e.g. their cardiovascular fitness, endurance, etc.) and can facilitate personalized prediction. See additionally pg. 1347, Section 3.3.2, “Context within and across activities”.); and providing the learned latent representation to the deterministic solver with the activity information (Morlock Fig. 1B, [0048] teaches applying each trained model to determining new data. Ni Fig. 3 shows the data flow from the attribute embedding layer and historical user sequences (providing the learned latent representation with the activity information) to the 2-layer stacked LSTM module of FitRec (the deterministic solver).)
Re. Claim 3, Ni/Morlock teaches the method of claim 2,
wherein the historical activity information comprises historical activity information for the user and includes physiological data associated with prior activities performed by the user prior to the future activity (Ni pg. 1346, Section 3.3 “Model Structure” teaches that the contextual embedding module and the attribute embedding layer encode user attributes and historical information into the embedded representation. Ni Fig. 3 and Table 2 teaches historical user sequences measurements include heart rate, timestamps, distance, speed.), and
wherein the prior activities do not include performance of the future activity (Ni Fig. 5 and pg. 1351 teaches instead of black box heart rate predictions given contextual inputs, providing personalized workout route recommendation… Given expected workout criteria (e.g. an expected workout time, an idiosyncratic heart rate or speed curve), suggesting routes (i.e. historical routes from all users) that will match the user’s expectation… if a user has particular heart rate targets in mind, or simply wishes to identify alternate routes that are ‘similar to’ (in terms of heart rate profile) a route a user has taken previously (do not include performance of the future activity)… could be that the user is traveling but wants to maintain their regular exercise routine. The Specification at para. 0038 describes “future workout information may include… (e.g., variations of a running… route)… for which the user is predicted to achieve a desired… heartrate, workout time, or other activity level… future workout information may be information obtained from other users… and/or from map data or other stored data (e.g., user-agnostic data) describing a future workout”. See alternately Ni pg. Figs. 4-5 and pg. 1350, Section 4.4 “Results: Workout profile forecasting”.)
Re. Claim 4, Ni/Morlock teaches the method of claim 2, wherein the historical activity information further comprises historical activity information for another user different from the user (Ni pg. 1349, Section 4.2 “Training Procedure” teaches sorting each user’s workouts (historical activity information) in chronological order based on the first timestamp of each workout… splitting and aggregating users’ first 80% of workouts for training, the next 10% for validation and the final 10% for testing (comprises historical activity information for another user).)
Re. Claim 6, Ni/Morlock teaches the method of claim 1, wherein the future activity comprises a workout, and wherein the activity information comprises workout parameters (Ni pg. 1344, Section 1. “Introduction” teaches FitRec takes as input… multiple workout measurement sequences (workout parameters)… FitRec can then be applied to (1) Quantitative tasks, such as personalized sequential modeling to predict how workout measurements (e.g. heart rate) will change across a workout (the future activity comprises a workout), either beforehand (i.e., based on a map of the intended route) or in real time as the user exercises. Additionally, Ni pg. 1351, Section 5 “Personalized Recommendation” teaches, given workout criteria (the workout parameters), suggesting routes that will match the user’s expectation.)
Re. Claim 7, Ni/Morlock teaches the method of claim 6, wherein the physiological prediction comprises a predicted heartrate zone for the user during the workout (Ni Fig. 2 and pg. 1344, Section 1. “Introduction” teaches applying FitRec to… predict how workout measurements (e.g. heart rate) will change… in real time as the user exercises (during the workout)… accounting for differences in comfort heart rate zones requires robust modeling techniques to capture the dynamics of user-dependent features. Ni Fig. 2 and pg. 1346, Section 3.1.2 “Short term prediction” teaches predicting the short term heart rate dynamics (comprising a predicted heartrate zone) during the activity in the next few seconds; and pg. 1351, Section 5 “Personalized Recommendation” teaches predicting whether the user’s heart rate will exceed some threshold (for the zone).)
Re. Claim 8, Ni/Morlock teaches the method of claim 6, wherein the physiological prediction comprises a predicted heartrate for the user during the workout (Ni pg. 1344, Section 1. “Introduction” teaches applying FitRec to (1) Quantitative tasks, such as personalized sequential modeling to predict how workout measurements (e.g. heart rate) will change across a workout (the physiological prediction for the user), either beforehand (i.e., based on a map of the intended route) or in real time as the user exercises (during the workout). Ni Fig. 3, pg. 1345, Section 3.1 “Notation and Tasks” and pg. 1346, Section 3.1.2 “Short term prediction” teaches predicting the target output yT (comprises a predicted heart rate for the user) at time step T.)
Re. Claim 9, Ni/Morlock teaches the method of claim 6, wherein the physiological prediction comprises a predicted number of calories that will be burned by the user during the workout (see claim 6 prior art rejection. Morlock Fig. 1 and [0157] teaches each network is trained to predict the desired output: traversal time and caloric expenditure for each segment.)
Re. Claim 10, Ni/Morlock teaches the method of claim 6, wherein the physiological prediction comprises a prediction of a potential cardiovascular event for the user during the workout (Ni pg. 1351, Section 5 “Personalized Recommendation” teaches predicting whether the user’s heart rate will exceed some threshold (a potential cardiovascular event for the user). Additionally, Ni pg. 1346, Section 3.2 “Relation Between the Two Tasks” teaches short term prediction is useful in scenarios like anomaly detection and real-time decision making (e.g. advising a user that they should slow down in the next minute to avoid exceeding their desired maximum heart rate).)
Re. Claim 11, Ni/Morlock teaches the method of claim 6, further comprising training the first neural network of the hybrid machine learning model (see claims 1 and 6 prior art rejections (a trained first LSTM). Additionally, Morlock Fig. 1A, [0047], [0086] teaches training each neural network and determining weighting factors and biases for prediction of travel time per trail segment 110 (the trained first neural network).) as one of
a user-demand model (Ni pg. 1344, Section I “Introduction” teaches the LSTM-based model FitRec considers context both within and across workouts… infers static user embeddings from user attributes (a user-demand model). Ni pg. 1346, Section 3.3 “Model Structure” teaches a basic structure… that encodes user attributes and historical information (a user-demand model) into an embedded representation. Morlock Fig. 1A step 104, [0154]-[0155] teaches creating several variations of averaged or integrated sensor measurements over the length of each segment that can be input into an artificial neural network, e.g., starting heart-rate, average heart rate, minimum heart rate, total heartbeats per trail segment, average blood oxygen level, maximum blood oxygen level… the amount of calories expended approximated by an index based on heart rate over time and pulse oximetry (user-demand). Alternately, Ni pg. 1345, Section 2 “Related Work” teaches personalized fitness recommendation that considers a variety of targets (user-demands) such as user preferences, goals, and environment.),
a fatigue model (Ni pg. 1344, Section I “Introduction” teaches… and meanwhile learns temporal embeddings (a fatigue model) from users’ recent workout sequences. Ni pg. 1346, Section 3.3 “Model Structure” teaches… the embedded representation captures… e.g. their cardiovascular fitness, endurance, etc. (a fatigue model). Morlock Fig. 1A steps 100-102, [0151]-[0152] teach measuring fundamental quantitative and qualitative indicators of fitness for several participants of the same activity such as minimal and maximal heart rate, pulse oximeter readings, body mass index, height, weight, weight of equipment, and age… collecting data for the above participants during traversal of several trail segments of varying degrees of difficulty, e.g., sensors used to collect geographic location and elevation, heart-rate, 3-axis acceleration, pulse oximetry/measures of oxygen content of the blood (fitness-fatigue).), or
a weather-demand model (Morlock Fig. 1A, [0153] teach collecting weather and other environmental data… that may influence the time to traverse a trail segment or the energy consumed (weather-demand). Additionally, Morlock [0134]-[0136] teaches the factors needed to predict traversal time and caloric burn, based on predetermined costs (demand) for a given activity, include Participant Ability: Level of Effort (user); Fitness (fitness-fatigue); Strength; Agility; Endurance/Stamina (fitness-fatigue); Modifiers—Weather (weather).)
Re. Claim 12, Ni/Morlock teaches the method of claim 1, further comprising: providing environmental information for the future activity to the hybrid machine learning model (Ni’s Fig. 3 Model structure), wherein the physiological prediction (Ni’s Fig. 3 outputs) is based in part on the environmental information (see claim 1 prior art rejection. Ni pg. 1345, Section 2 “Related Work” teaches personalized fitness recommendation that considers a variety of targets such as user preferences, goals, and environment. Morlock [0134]-[0136], [0139] teaches the factors needed to learn costs and predict the traversal time and the caloric burn include weather information (environmental information). Also, Morlock Fig. 1A, [0153] teaches collecting weather and other environmental data… that may influence the time to traverse a trail segment or the energy consumed.)
Re. Claim 13, Ni/Morlock teaches the method of claim 12, wherein the environmental information comprises a location of the future activity, a temperature, a humidity, or other weather information (Ni pg. 1344, Section I, “Introduction” teaches evaluating FitRec on workout profile forecasting, such as estimating a user’s likely speed and heart rate profile given the activity they intend to perform (e.g. cycling a particular GPS route) (location). Morlock Fig. 1A, [0153] teaches collecting weather and other environmental data such as precipitation, wind speed/direction, temperature ground cover… that may influence the time to traverse a trail segment or the energy consumed. Morlock [134] teaches if you know how the weather (other weather information) influences the energy requirement, then the time required to traverse a trail and the energy output needed can be predicted. Morlock [0135]-[0136] teaches real-time and historical weather as prediction factors.)
Re. Claim 14, the subject matter of claim 14 is essentially defined in terms of a system, which is technically corresponding to claim 1. Since claim 14 is analogous to claim 1, it is similarly analyzed and rejected in a manner consistent with the rejection of claim 1. Further, Ni/Morlock teaches a device, comprising: a memory; and one or more processors configured to perform the method according to claim 1. (Morlock Fig. 2, [0181]-[0182], [0185], [0232] teaches a personal portable training device 200, which includes a processor 202 operatively coupled to a memory 220 resource and is configured to acquire manual input, communicate with one or more servers, and perform calculations locally on the device.)
Re. Claim 15, the subject matter of claim 15 is essentially defined in terms of a system, which is technically corresponding to claim 2. Since claim 15 is analogous to claim 2, it is similarly analyzed and rejected in a manner consistent with the rejection of claim 2.
Re. Claim 19, the subject matter of claim 19 is essentially defined in terms of a manufacture, which is technically corresponding to claim 1. Since claim 19 is analogous to claim 1, it is similarly analyzed and rejected in a manner consistent with the rejection of claim 1. Further, Ni/Morlock teaches a non-transitory machine-readable medium comprising code that, when executed by a processor, causes the processor to perform the method according to claim 1 (Morlock Fig. 2, [0181]-[0182], [0185], [0240] teaches a personal portable training device 200, which includes a processor 202 operatively coupled to a memory 220 resource… software may be stored in the form of a machine-readable storage device… that are suitable for storing program(s) that, when executed, implement the embodiments of the present invention.)
Re. Claim 20, the subject matter of claim 20 is essentially defined in terms of a manufacture, which is technically corresponding to claim 2. Since claim 20 is analogous to claim 2, it is similarly analyzed and rejected in a manner consistent with the rejection of claim 2.
Re. Claim 22, Ni/Morlock teaches the non-transitory machine-readable medium of claim 19, wherein the future activity comprises a workout (see claim 6 prior art rejection), wherein the activity information comprises workout parameters (see claim 6 prior art rejection), and wherein the physiological prediction comprises at least one of a predicted heartrate zone for the user during the workout (see claim 7 prior art rejection), a predicted heartrate for the user during the workout (see claim 8 prior art rejection), a predicted number of calories that will be burned by the user during the workout (see claim 9 prior art rejection), and a prediction of a potential cardiovascular event for the user during the workout (see claim 10 prior art rejection).
Re. Claim 23, Ni/Morlock teaches the method of claim 1, wherein the user input comprises a user selection of the future activity (Morlock Fig. 5 teaches the processor selects a route at random or allows the user to select from a list of available routes), and wherein the output comprises one or more user-specific predictions for the future activity (Ni pg. 1351, Section 5 “Personalized recommendation” teaches going from ‘black box’ heart rate predictions (see examples in Fig. 4) with FitRec to application of FitRec in practice to provide a personalized workout route recommendation (prediction output) given expected workout criteria / the user’s expectation, such as particular heart rate targets. Ni Fig. 5 teaches examples of a recommended route with comparable hill-climbs.)
Re. Claim 24, Ni/Morlock teaches the method of claim 1, wherein the user input comprises a user physiological goal, and wherein the output comprises a recommended workout corresponding to the user physiological goal (Ni pg. 1351, Section 5 “Personalized recommendation” teaches providing personalized workout route recommendation given expected workout criteria / the user’s expectation, such as particular heart rate targets (user physiological goals). Morlock Fig. 5 teaches a user data entry step.)
Claims 17-18 are rejected under 35 U.S.C. 103 as being unpatentable over Ni in view of Morlock; Shelton, IV et al. (US 2022/0238216 A1; “Shelton” herein); “The Use of Fitness-Fatigue Models for Sport Performance Modelling: Conceptual Issues and Contributions from Machine-Learning” (3/3/2022) to Imbach et al. (“Imbach” herein); and “Metabolic reactions to work in the desert” (1967) to Klausen et al. (“Klausen” herein).
Re. Claim 17, Ni/Morlock teaches the device of claim 15,
wherein the trained first neural network of the hybrid machine learning model (see claim 1 prior art rejection) further comprises a user-demand model that translates an instantaneous activity intensity […] (see claim 11 prior art rejection (a trained user-demand model). Ni Fig. 2 and pg. 1344, Section I “Introduction” teaches forecasting what the heart rate profile (translates instantaneous activity intensity) will look like / how it will change across a workout based on a map of the intended route.), and
wherein the hybrid machine learning model (see claim 1 prior art rejection) further comprises a trained third neural network trained as a fatigue model that describes […] (see claim 11 prior art rejection (a trained fatigue model). Morlock Fig. 1A steps 100-102, [0151]-[0152] teach collecting data for the above participants during traversal of several trail segments of varying degrees of difficulty, e.g., sensors used to collect geographic location and elevation, heart-rate, 3-axis acceleration, pulse oximetry/measures of oxygen content of the blood. Morlock Fig. 5 teaches route calculation based on a requested of level of effort), and a trained fourth neural network trained as a weather-demand model that describes […] (Morlock Fig. 1A, [0153] teach collecting weather and other environmental data… that may influence the time to traverse a trail segment or the energy consumed (a trained weather-demand model). Morlock [0134]-[0136] teaches the factors needed to predict traversal time and caloric burn, based on predetermined costs (demand) for a given activity, include… Modifiers—Weather.)
Ni/Morlock may not teach
a user-demand model that translates an instantaneous activity intensity into an oxygen demand of the user for that intensity;
a fatigue model that describes fatigue incurred over time during the future activity; or
a weather-demand model that describes a change in oxygen demand of the user as a function of one or more weather parameters.
Shelton teaches
translates an instantaneous activity intensity into an oxygen demand of the user for that intensity ([0050]-[0051] teaches the VO2 max sensing system may select correct VO2 max data measurements including heart rate measurements (Ni’s predicted heart rate profile) during correct time segments to calculate accurate VO2 max information (translate… into oxygen demand).);
Therefore, it would have been prima facie obvious to one of ordinary skill in the art before the effective filing date to have modified the method of modeling heart rate and activity data for personalized fitness recommendation of Ni/Morlock to calculate accurate VO2 max information (i.e., a predicted VO2 max profile) using the predicted heart rate profile and to use this information as part of a computing system applying machine learning to a data collection as taught by Shelton, with the motivation of improving data from wearables, improving artificial intelligence algorithm iterations and improving patient outcomes (see Shelton at Title and para. 0003-0005).
Imbach teaches
describes fatigue incurred over time during the future activity (Abstract teach that in Fitness-Fatigue impulse responses models (FFMs) (modeling fatigue incurred over time, see equation 1 and pg. 3, 1st para.), athletic performance is described by first order transfer functions which represent Fitness and Fatigue antagonistic responses to training… Weaknesses of FFMs may be surpassed by embedding physiological representation of training effects into non-linear and multivariate learning algorithms… the machine-learning approach is a way to take advantage of models based on physiological assumptions within powerful machine-learning models.)
Therefore, it would have been prima facie obvious to one of ordinary skill in the art before the effective filing date to have modified the method of modeling heart rate and activity data for personalized fitness recommendation of Ni/Morlock/Shelton to make exercise predictions using fitness-fatigue and ML models (see Imbach at Fig. 1) and to use this information as part of a method using fitness-fatigue models for sport performance modelling as taught by Imbach, with the motivation of improving predictive capabilities for investigation of the relationships between exercise training and performance (see Imbach at pg. 1, Abstract and Key Points).
Klausen teaches
describes a change in oxygen demand of the user as a function of one or more weather parameters (Fig. 1 and pg. 293, “Results” teach the oxygen uptake (change in oxygen demand) at various submaximal workloads in four subjects during desert experiments (air temperature 32.1 C) and during control experiments in comfortable environments (air temperature 24.8 C) (as a function of one or more weather parameters). See also Figs. 3-4.)
Therefore, it would have been prima facie obvious to one of ordinary skill in the art before the effective filing date to have modified the method of modeling heart rate and activity data for personalized fitness recommendation of Ni/Morlock/Shelton/Imbach to perform data collection and analysis of the relationship between oxygen uptake and human work performance at different air temperatures and to use this information as part of methods of study of metabolic reactions to work in the desert as taught by Klausen, with the motivation of improving human understanding of the effect of hot, dry environments on an individual’s work performance (see Klausen at pg. 292, first para. after Abstract and keywords).
Re. Claim 18, Ni/Morlock/Shelton/Imbach/Klausen teaches the device of claim 17, wherein the physiological state equation (see claim 1 prior art rejection) further comprises the trained fatigue model, and the trained weather-demand model as functions within the physiological state equation (Ni pg. 1347, Section 3.3.3 “FitRec” teaches equation (6). Ni pg. 1347, Section 3.3.2 “Context within and across activities” teaches learning attribute embeddings and contextual embeddings for each attribute input; see equations (3)-(5) (trained models used as functions within the physiological state equation). See claim 17 prior art rejection. Imbach at Abstract, equation 1, and pg. 3, 1st para. teaches fitness and fatigue impulse responses and embedding physiological representation of training effects into non-linear and multivariate learning algorithms (describes the trained fatigue model). Klausen Fig. 1 and pg. 293, “Results” teach the oxygen uptake at various submaximal workloads in four subjects during desert experiments (air temperature 32.1 C) and during control experiments in comfortable environments (air temperature 24.8 C) (describes the trained weather-demand model).)
Claim 21 is rejected under 35 U.S.C. 103 as being unpatentable over Ni in view of Morlock and Liu et al. (2019) (“Predicting the Heart Rate Response to Outdoor Running Exercise”; “Liu” herein).
Re. Claim 21, Ni/Morlock teaches the non-transitory machine-readable medium of claim 19, wherein generating the physiological prediction for the user (see claim 1 prior art rejection) comprises generating the physiological prediction for the user based at least in part on a representation for the user, the representation generated using [… a Gaussian process…] (Ni pg. 1346, Section 3.3 “Model Structure” teaches both models share the same basic structure consisting of a contextual embedding module and an attribute embedding layer. These layers encode user attributes and historical information into an embedded representation (a representation for the user) that captures their latent individual attributes (e.g. their cardiovascular fitness, endurance, etc.) and can facilitate personalized prediction. See additionally Ni pg. 1347, Section 3.3.2, “Context within and across activities”. The Examiner notes it is unclear what a Gaussian process must or must not entail.)
Ni may not teach a Gaussian process.
Liu teaches a Gaussian process (pg. 219, Section III.B. “Parameter estimation” teaches defining the optimal parameters as the values that minimize the cost function F(a). Levenberg-Marquardt is adopted to find the parameters that minimize F(a). Levenberg-Marquardt algorithm is a standard technique… a combination of the gradient descent algorithm and Gauss-Newton algorithm.)
Therefore, it would have been prima facie obvious to one of ordinary skill in the art before the effective filing date to have modified the method of modeling heart rate and activity data for personalized fitness recommendation of Ni/Morlock to define optimal parameters using a standard technique with a Gauss-Newton algorithm and to use this information as part of a method predicting the heart rate response to outdoor running exercise as taught by Morlock, with the motivation of improving the accuracy and ease of health exercise monitoring and improving heart rate prediction modeling (see Liu at Abstract; pg. 217, Section I “Introduction”; and pg. 218, Section II “Background and related work”).
Response to Arguments
Claim Objections
Regarding objection to claim 17, claim 17 has been amended to obviate the previous claim objection, hereby withdrawn.
Rejection under 35 U.S.C. §103
Regarding the rejection of Claims 1-4, 6-15 and 17-24, the Examiner has considered the Applicant’s arguments but does not find them persuasive for at least the following reasons:
Re. Reason 1 (“The Office Action does not suggest that the proposed combination of Ni and Morlock discloses or suggests, “determining, with the deterministic solver, a deterministic solution to the physiological state equation that includes the trained first neural network as the function”, as recited in claim 1 (emphasis added)”) in remarks at pgs. 9-10:
The Examiner respectfully submits that, given the broadest reasonable interpretation in light of the specification, Ni in view of Morlock teaches or renders obvious the claimed features. The generating step previously discussed includes discussion of this deterministic solution (the physiological prediction for the user with respect to the future activity) that is determined with the solver (software module, i.e., 2-layer stacked LSTM module, that applies data to a machine learning algorithm and reports the results). The deterministic solution to the physiological state equation is mapped to the generated outputs of Ni, which are shown in Fig. 3 (colored in blue).
Re. Reason 2 in remarks at pg. 10: The recitation of a “deterministic solver” does not connote structure. This “deterministic solver” given the broadest reasonable interpretation in light of the specification can be mapped to any software instructions providing the functionality of solving “a physiological state equation that includes the trained first neural network as a function within the physiological state equation”. In computer science, a deterministic solver or algorithm is an algorithm that, given a particular input, will always produce the same output. The FitRec software module of Ni applies data to an algorithm and an output is generated. Ni’s software module is used as designed to supply data and output the result.
The specification at para. 0039 states: "In one or more other implementations, the solver 302 may generate a deterministic solution to the PSE 304. For example, in one or more implementations, the solver 302 may solve the PSE 304 using an iterative operation (e.g., a Fourth Order Runge-Kutta method) to generate the physiological prediction(s) responsive to receiving the embedding, z, the future workout information, and/or the environmental information (e.g., such that the PSE 304 yields an solution that is differentiable against its input parameters). In other implementations, the solver 302 may be implemented as a neural network trained to generate the physiological prediction(s) responsive to receiving the embedding z, the future workout information, and/or the environmental information" (emphasis added).
The claimed invention does not recite the deterministic solver solving the PSE using an iterative operation, such as a Fourth Order Runge-Kutta method. Applicant has chosen to claim a different implementation. It is understood that the implementation that is chosen to solve the PSE will “generate a deterministic solution to the PSE 304”. The claims recite the deterministic solver as being implemented to solve a PSE that includes a trained neural network as a function within the PSE. Ni has been mapped to this limitation appropriately as Ni in view of Morlock render obvious the use of a portable electronic device (Morlock’s portable training device 200) providing a hybrid machine learning model (Ni’s Model structure of Fig. 3) that includes a trained first neural network (Ni’s trained first LSTM seen in Fig. 3) and a deterministic solver (Ni’s “2-layer stacked LSTM module” implementing the Model structure of Fig. 3), wherein the deterministic solver (Ni’s “2-layer stacked LSTM module”) does not connote structure and is a subroutine providing the functionality of processing a physiological state equation (Ni’s “2-layer stacked LSTM model”) that includes the trained first neural network as a function within the PSE (Ni’s exemplary LSTM seen within the 2-layer stacked LSTM model for workout profile forecasting that is implemented according to its own subroutine). As such, the rejection of the independent claims with the combination of Ni and Morlock is respectfully maintained
Re. Reason 3 in remarks at pg. 10-11, the Examiner respectfully asserts for reasons previously given that Ni does teach “a deterministic solver” given the broadest reasonable interpretation in light of the specification. The deterministic solver does not connote structure and is interpreted as a software subroutine that includes the PSE functionality that includes a trained neural network. Ni in view of Morlock teaches implementing Ni’s subroutine on Morlock’s portable training device. Ni’s 2-layer stacked LSTM module is a subroutine specific to a version of FitRec for workout profile forecasting (as opposed to the subroutine for short term prediction).
Re. Reason 4 in remarks at pg. 11, the Examiner respectfully asserts for reasons previously given that Ni does teach “a physiological state equation” given the broadest reasonable interpretation in light of the specification. The deterministic solver does not connote structure and is interpreted as a software subroutine that includes the PSE functionality that includes a trained neural network. Ni in view of Morlock teaches implementing Ni’s subroutine on Morlock’s portable training device. Ni’s 2-layer stacked LSTM model implemented by the associated subroutine for workout profile forecasting reads on the recited PSE.
Re. Reason 5 in remarks at pg. 11-12, the Examiner respectfully asserts for reasons previously given that Ni does teach “a hybrid machine learning model” given the broadest reasonable interpretation in light of the specification. Ni has been mapped to this limitation appropriately as Ni in view of Morlock render obvious the use of a portable electronic device (Morlock’s portable training device 200) providing a hybrid machine learning model (Ni’s Model structure of Fig. 3) that includes a trained first neural network (Ni’s trained first LSTM seen in Fig. 3 in the workout profile forecasting model) and a deterministic solver (Ni’s “2-layer stacked LSTM module” implementing the workout profile forecasting model). The Examiner respectfully asserts that “hybrid machine learning model” is not a term of art; and all the parts required to map to a hybrid machine learning model are all found in the basis of rejection.
Note: The claim mapping with Morlock adds more evidence that trained neural networks are taught in the prior art and were previously known to be trained based on recorded physiological data for past activities. The hybrid machine learning model is a sum of its parts.
Regarding the rejection of Claims 2-4, 6-15 and 17-24, the Applicant has not offered any arguments with respect to these claims other than to reiterate the argument(s) present for analogous claim 1 or for the independent claim from which they depend. As such, the rejection of these claims is also maintained.
Rejections under 35 U.S.C. §101
Note: As the USPTO still has not trained Examiners on the Ex Parte Desjardins memorandum, it remains unclear what Examiners are supposed to differently in prosecution. Examiners already examine cases in accordance with the MPEP and memorandums incorporated therein. Previous memorandums appear to provide the substance of the Ex Parte Desjardins memorandum.
Regarding the rejection of Claims 1-4, 6-15 and 17-24, the Examiner has considered the Applicant’s arguments but does not find them persuasive for at least the following reasons:
Re. Reason 1 in remarks at pg. 15, the Examiner respectfully submits that the additional element of the hybrid machine learning model (executing at the portable electronic device) is analyzed in the subject matter eligibility analysis steps 2A2 and 2B. This model executes (i.e., “implements”) the identified abstract idea (e.g., “generating… a physiological prediction…”) at a high-level of generality and is merely invoked as a tool to perform an existing process, such that this amounts no more than mere instructions to apply the abstract idea using a general-purpose computer (i.e., the “portable electronic device” previously considered in subject matter eligibility Steps 2A2 and 2B). “Even in combination, these additional elements…” (i.e., the hybrid machine learning model executing on the portable electronic device) “do not integrate the abstract idea into a practical application…” Also, “mere instructions to apply an exception using a generic computer…” (i.e., on the portable electronic device) “cannot provide an inventive concept.” Accordingly, alone or in combination, the additional element of the hybrid machine learning model (executing on the portable electronic device) does not provide significantly more.
Re. Reason 2 in remarks at pgs. 15-17, Applicant argues that this discussion is not about reducing the computational load of a cloud-based server (or any other technical improvement to a cloud-based server) by allowing distributed processing. This makes sense as Applicant has not recited a cloud-based server. In that case, Applicant is understood to be asserting that the portable electronic device improves the ability of the portable electronic device to output feedback by improving the computing functionality of the portable electronic device.
First, the Examiner respectfully submits that an improvement to the abstract idea of outputting feedback by executing the hybrid machine learning model on the portable electronic device does not provide a practical application or an inventive concept. Mere instructions to apply the (improved) abstract idea on a generic computer does not provide a practical application or an inventive concept.
Considering the second part further, as the Applicant is asserting that the invention improves upon conventional functioning of the portable electronic device, the Examiner respectfully submits that the provided evidence from the Applicant’s disclosure is a mere conclusory statement that does not even specify what computer functionality is alleged as improved. The specification sets forth an improvement to “computer functionality”; however, the explanation present in the specification does not explicitly set forth an improvement to “computational and/or power resources” available to the portable electronic device. The specification merely states that an end user device may generally have less computational power and/or resources than a cloud-based server. Regardless of what the alleged “computer functionality” may be, the specification presents a bare assertion of an unspecified improvement and fails to provide sufficient details as to how the portable electronic device may be improved. MPEP § 2106.05(a).
Re. Reason 3 in remarks at pgs. 17-18, the Examiner respectfully submits that the hybrid machine learning model has been considered alone and in combination with the portable electronic device for subject matter eligibility analysis steps 2A2 and 2B. See Examiner’s response to reason 1. Further, the Examiner respectfully submits that the claims, as a whole, do not improve the computing functionality of the claimed portable electronic device for at least the reasons provided in Examiner’s response to reason 2. Given the broadest reasonable interpretation, the claims as drafted recite the abstract idea of a mental process alone or along with a mathematical concept (step 2A1 = YES). The additional elements are considered but these do not provide an integration of the judicial exception into a practical application (step 2A2 = NO). Mere instructions to apply the judicial exception using generic computers and generic computer components cannot provide an inventive concept; also, the use of a hybrid machine learning model or algorithm in its ordinary capacity to perform tasks in the identified abstract idea does not provide an inventive concept, even in combination with generic computers and generic computer components (step 2B = NO). Thus, the claims are subject matter ineligible.
(1) The Examiner respectfully submits that (a) the trained first neural network that has been trained based at least in part on prior physiological data associated with prior activities performed prior to the future activity; and (b) the deterministic solver for a physiological state equation that includes the trained first neural network as a function within the physiological state equation are part of the identified abstract idea and are not part of the additional element(s). For example, the trained neural network is already trained and is a function (and is recited as such in the claim). The deterministic solver is a software subroutine that includes the trained neural network as a function. The mathematical models are part of the instructions to apply the judicial exception with a generic machine learning algorithm (“hybrid machine learning model”) to perform an existing computational process on a generic computer.
In computer science, a deterministic algorithm is an algorithm that, given a particular input, will always produce the same output. The Examiner respectfully submits that the plain meaning of a deterministic solver is inconsistent with the Applicant’s disclosure, which does not describe a “deterministic” solver, and which describes the claimed solver as not being an implementation that would render a deterministic solution.
[0039] states: "In one or more other implementations, the solver 302 may generate a deterministic solution to the PSE 304. For example, in one or more implementations, the solver 302 may solve the PSE 304 using an iterative operation (e.g., a Fourth Order Runge-Kutta method) to generate the physiological prediction(s) responsive to receiving the embedding, z, the future workout information, and/or the environmental information (e.g., such that the PSE 304 yields an solution that is differentiable against its input parameters). In other implementations, the solver 302 may be implemented as a neural network trained to generate the physiological prediction(s) responsive to receiving the embedding z, the future workout information, and/or the environmental information".
The claimed invention does not recite the solver solving the PSE using an iterative operation (e.g., a Fourth Order Runge-Kutta method), but the claimed invention does recite the solver being implemented as a neural network trained to generate a physiological prediction responsive to receiving user input. Without a deterministic algorithm disclosure, it is unclear what the word "deterministic" adds to the claim limitation. Without reciting an iterative operation, it is unclear what “deterministic solver” must or must not entail. And by reciting a trained neural network, it appears that the algorithm may not be deterministic, such that it is unclear how computational resources on the portable electronic device are being spared by the claimed invention.
(2) Executing the hybrid machine learning model at the portable electronic device amounts no more than mere instructions to apply the abstract idea using a general-purpose computer. Even in combination, these additional elements do not integrate the abstract idea into a practical application; and mere instructions to apply an exception on the portable electronic device cannot provide an inventive concept. Accordingly, alone or in combination, the additional element of the hybrid machine learning model (executing on the portable electronic device) does not provide significantly more.
Integral use of a machine to achieve performance of a method may integrate the recited judicial exception into a practical application or provide significantly more, in contrast to where the machine is merely an object on which the method operates, which does not integrate the exception into a practical application or provide significantly more. See CyberSource v. Retail Decisions, 654 F.3d 1366, 1370, 99 USPQ2d 1690, 1694 (Fed. Cir. 2011) ("We are not persuaded by the appellant's argument that the claimed method is tied to a particular machine because it ‘would not be necessary or possible without the Internet.’ . . . Regardless of whether "the Internet" can be viewed as a machine, it is clear that the Internet cannot perform the fraud detection steps of the claimed method"). For example, as described in MPEP 2106(f), additional elements that invoke computers or other machinery merely as a tool to perform an existing process will generally not amount to significantly more than a judicial exception. See, e.g., Versata Development Group v. SAP America, 793 F.3d 1306, 1335, 115 USPQ2d 1681, 1702 (Fed. Cir. 2015) (explaining that in order for a machine to add significantly more, it must "play a significant part in permitting the claimed method to be performed, rather than function solely as an obvious mechanism for permitting a solution to be achieved more quickly"). MPEP 2106.05(b)(II).
In this case, the hybrid machine learning model acts is merely an object on which the abstract idea applies the mathematical functions to input data and output output data on a generic computer. The Examiner asserts that the hybrid machine learning model invokes a computer merely as a tool to perform mathematical calculations as part of the abstract idea. The Examiner asserts that this use of the additional element to apply the judicial exception on the computer does not amount to significantly more than the judicial exception as it does not play a significant part in permitting the computer to perform the method with a machine learning algorithm. The hybrid machine learning model functions solely as an obvious mechanism (a generic machine learning algorithm) for applying data and reporting the results quicker than a computer can do without such a mechanism.
Note: In McRO, the Federal Circuit held claimed methods of automatic lip synchronization and facial expression animation using computer-implemented rules to be patent eligible, because they were not directed to an abstract idea. McRO, 837 F.3d at 1316, 120 USPQ2d at 1103. The basis for the McRO court's decision was that the claims were directed to an improvement in computer animation and thus did not recite a concept similar to previously identified abstract ideas. Id. MPEP 2106.05(a). The Examiner respectfully submits that the Applicant has not invented a new machine learning algorithm. Also, the court relied on the specification's explanation of how the claimed rules enabled the automation of specific animation tasks that previously could not be automated. 837 F.3d at 1313, 120 USPQ2d at 1101. The Examiner asserts that the claims do not recite a technical solution to a technical problem and that the Applicant’s disclosure does not appear to describe a technical solution to a technical problem related to the machine learning technology or the portable electronic device. The operation of either of these devices is not improved.
Re. Reason 4 in remarks at pgs. 18-20, the Examiner respectfully reminds the Applicant that the identified claim elements of “a trained neural network” (trained outside the scope of the claim) (i.e., a mathematical function) and “determining, with the deterministic solver, a deterministic solution to the physiological state equation that includes the trained first neural network as the function” within the physiological state equation (i.e., solving, with the deterministic solver (which does not connote structure), the physiological state equation in this manner), when given broadest reasonable interpretation in light of the disclosure, encompass the aforementioned mathematical concept. As such, the Examiner respectfully asserts that this judicial exception cannot provide a practical application. MPEP 2106(I). As the Application of the trained neural network and the solving of the physiological state equation represent mathematical concepts, they need not be practically performed in the human mind. Such is a requirement of a separate abstract idea sub-grouping.
Note: The Examiner additionally refers the Applicant to the July 2024 Subject Matter Eligibility Guidance Example 47. This example relates to anomaly detection and illustrates the application of the eligibility analysis to claims that recite limitations specific to artificial intelligence, particularly the use of an artificial neural network (a machine learning model like Applicant’s hybrid machine learning model). Example Claim 2 is ineligible because it recites a judicial exception (abstract idea), and the claim as a whole does not integrate the exception into a practical application (and is thus directed to an abstract idea), and the claim does not provide significantly more than the exception (does not provide an inventive concept).
Example Claim 2 recites a method of using an artificial neural network (ANN) comprising: (a) receiving, at a computer, continuous training data;
(b) discretizing, by the computer, the continuous training data to generate input data;
(c) training, by the computer, the ANN based on the input data and a selected training algorithm to generate a trained ANN, wherein the selected training algorithm includes a backpropagation algorithm and a gradient descent algorithm;
(d) detecting one or more anomalies in the data set using the trained ANN;
(e) analyzing the one or more detected anomalies using the trained ANN to generate anomaly data; and
(f) outputting the anomaly data from the trained ANN.
In this example, step (d) recites detecting one or more anomalies in a data set using the trained ANN. The claim does not provide any details about how the trained ANN operates or how the detection is made, and the plain meaning of “detecting” encompasses mental observations or evaluations, e.g., a computer programmer’s mental identification of an anomaly in a data set.
Also, example step (e) recites analyzing the one or more detected anomalies using the trained artificial neural network to generate anomaly data. The step of analyzing includes both determining that an anomaly has been detected and may further include suggesting a type or cause of the anomaly. The plain meaning of “analyzing” encompasses evaluating information, which in this claim is limited to evaluating detected anomalies to generate anomaly data by the trained ANN. The claim does not limit how the analysis (evaluation) is performed, and there is nothing about a detected anomaly itself that would limit how it can be analyzed. As explained in the background, “the anomaly data may explain the type of anomaly or a cause of the anomaly.” The claim does not include any additional details that explain the analysis of detected anomalies.
Applicant’s Claim 1 recites providing activity information for the future activity to the deterministic solver of the hybrid machine learning model (responsive to receiving the user input corresponding to the future activity) and generating, based on the provided activity information, a physiological prediction for the user with respect to the future activity by determining, with the deterministic solver, a deterministic solution to the physiological state equation that includes the trained first neural network (i.e., a function) as the function. The claim does not provide any details about (1) how the portable electronic device operates, how data is received or (2) how the hybrid machine learning model operates, how data is provided, or how data is generated. Also, the plain meaning of “receiving”, “providing” and “generating” encompasses mental observations or evaluations that can be performed with the human eye and mental observations or evaluations that can be output on paper with the aid of a pen, e.g., a health sciences computer programmer’s mental reception of information and generation of a physiological prediction for the user with respect to the future activity. Further, the claim does not limit how the receiving (observation), providing (observation), or generating (evaluation) is performed. As such, (1) there is nothing about a user input corresponding to the future activity itself that would limit it can be received or provided; and (2) there is nothing about a physiological prediction itself that would limit how it can be generated.
Furthermore, the example discusses how step (f) for outputting the anomaly data merely requires a generic output using the trained ANN. The claim does not impose any limits on how the data is output or require any particular components that are used to output the anomaly data.
Applicant’s claim 1 recites providing an output corresponding to the generated physiological prediction in a manner that merely requires a generic output generated using the hybrid machine learning model and output using a generic computer. The claim does not impose any limits on how the data is output or require any particular (i.e., non-generic) components that are used to output the physiological prediction.
The Applicant’s claim 1 is comparable to Example 47 claim 2, which is not subject matter eligible. With this additional reasoning, it can be seen that the claimed invention of the instant application is subject matter ineligible.
Regarding the rejection of Claims 2-4, 6-15 and 17-24, the Applicant has not offered any arguments with respect to these claims other than to reiterate the argument(s) present for analogous claim 1 or for the independent claim from which they depend. As such, the rejection of these claims is also maintained.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure:
Mahyari et al. (“Physical Exercise Recommendation and Success Prediction Using Interconnected Recurrent Neural Networks”) for teaching the system, consisting of two inter-connected recurrent neural networks (RNNs), uses the history of workouts to recommend the next workout activity for each individual. The system then predicts the probability of successful completion of the predicted activity by the individual (see Abstract).
Neumann (US 2021/0162261) for teaching an exemplary embodiment 300 of computing device 104 providing a notification as a function of the human subject specific pattern… Computing device 104 may determine that a subject is engaging in a real-time human subject activity may be determined by a wearable device and/or sensor… wherein human subject activity 304 is transmitted to computing device 104 during a task, exercise, etc. A human subject activity 304 may be used as training data with a machine-learning model… while being collected by a computing device 104… Human subject specific pattern 308 may be a threshold determined by computing device 104, where a computing device 104 may retrieve a recommended healthy range of physiological state data… e.g., a safe heart rate range for a subject at particular age ranges, during particular exercises, with certain diseases, and the like. In such an example, when computing device 104 receive wearable fitness device 116 signals that indicate a heart rate above the healthy threshold, computing device 104 may alert a subject via a notification 208 that the human subject specific pattern 308 indicates the heart rate is above a healthy range, indicating exertion. See Fig. 3 and para. 0081.
Lyke et al. (US 2021/0093919 A1) for teaching workout recommendations that closely align with the user's traits. In one exemplary embodiment, workout data for a population of different individuals is analyzed to identify groups of similarly performing individuals. Each group of individuals is analyzed to generate an expected profile that approximates the physiological and/or psychological traits of the group. An expected profile includes heuristics and/or performance metrics that enable dynamic coaching during workouts (see Abstract).
THIS ACTION IS MADE FINAL. Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Jessica M Webb whose telephone number is (469)295-9173. The examiner can normally be reached Mon-Fri 9:00am-3:00pm CST.
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, Robert Morgan can be reached on (571) 272-6773. 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.
/J.M.W./Examiner, Art Unit 3683
/CHRISTOPHER L GILLIGAN/Primary Examiner, Art Unit 3683