Detailed Action
This Office Action is in response to the remarks entered on 03/24/2026. Claims 1-20 are currently pending.
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 .
Note
Examiner notes that if the applicant expands claims 1-10 and 11-15 in a manner that changes the scope of the claims from the original presented claimed inventive concept as presented in claims 16-20 in a way that is not a generic environmental implemental, the claims might be subject to restriction based on original presentation.
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 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Regarding claim 1,
2A Prong 1:
A method comprising: generating,
computing
combining, and (a mathematical concept – calculating the sum of the difference between the first vector and the second vector and the third vector)
generating, a digital content recommendation for the third segment of the client devices by decoding the third latent vector representation as combined with the change vector
2A Prong 2:
receiving, by a processing device via a network, input data describing interactions with digital content by client devices included in a group of client devices, respectively; (insignificant extra-solution activity MPEP 2106.05(g)(iii) of gathering statistics)
generating, by the processing device, … using an encoder of a machine learning model (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
computing, by the processing device (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
combining, by the processing device (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
generating, by the processing device, … using a decoder of the machine learning model (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
The additional elements as disclosed above alone or in combination do not integrate the judicial exception into practical application as they are mere insignificant extra solution activity, combination of generic computer functions that are restricted to field of use are implemented to perform the disclosed abstract idea above.
2B:
receiving, by a processing device via a network, input data describing interactions with digital content by client devices included in a group of client devices, respectively; (indicated as an insignificant extra-solution activity MPEP 2106.05(g)(iii) in Step 2A Prong 2. Therefore, the limitation is re-evaluated in Step 2B as well understood, routine and conventional activity MPEP 2106.05(d)(II)(i) of transmitting or receiving data over a network)
generating, by the processing device, … using an encoder of a machine learning model (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
computing, by the processing device (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
combining, by the processing device (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
generating, by the processing device, … using a decoder of the machine learning model (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
The additional elements as disclosed above in combination of the abstract idea are not sufficient to amount to significantly more than the judicial exception as they are well, understood, routine and conventional activity as disclosed in combination of generic computer functions and usage of elements that are restricted to field of use that are implemented to perform the disclosed abstract idea above.
Regarding claim 2,
2A Prong 1: Incorporates the rejection of claim 1.
2A Prong 2: wherein the machine learning model is trained on the input data without an indication of the digital content recommendation. (mere instructions to apply an exception using a computer MPEP 2106.05(f) – generic training process of a model)
2B: wherein the machine learning model is trained on the input data without an indication of the digital content recommendation. (mere instructions to apply an exception using a computer MPEP 2106.05(f) – generic training process of a model)
Regarding claim 3,
2A Prong 1: further comprising regularizing the latent space using a Kullback-Leibler divergence loss. (a mathematical concept – normalizing the latent space matrix (a set of values) using a math calculation)
2A Prong 2: This judicial exception is not integrated into a practical application.
2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
Regarding claim 4,
2A Prong 1: Incorporates the rejection of claim 1.
2A Prong 2: wherein the machine learning model is a variational autoencoder. (a field of use and technological environment MPEP 2106.05(h))
2B: wherein the machine learning model is a variational autoencoder. (a field of use and technological environment MPEP 2106.05(h))
Regarding claim 5,
2A Prong 1: Incorporates the rejection of claim 1.
2A Prong 2: wherein the input data includes categorical data and numerical data. (a field of use and technological environment MPEP 2106.05(h))
2B: wherein the input data includes categorical data and numerical data. (a field of use and technological environment MPEP 2106.05(h))
Regarding claim 6,
2A Prong 1: wherein the machine learning model is trained using a binary cross-entropy loss for the categorical data. (a mathematical concept – calculating difference between prediction output and ground truth)
2A Prong 2: This judicial exception is not integrated into a practical application.
2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
Regarding claim 7,
2A Prong 1: wherein the machine learning model is trained using a mean squared loss for the numerical data. (a mathematical concept - calculating difference between prediction output and ground truth)
2A Prong 2: This judicial exception is not integrated into a practical application.
2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
Regarding claim 8,
2A Prong 1: Incorporates the rejection of claim 5.
2A Prong 2: wherein the machine learning model processes the categorical data using a softmax activation. (a field of use and technological environment MPEP 2106.05(h))
2B: wherein the machine learning model processes the categorical data using a softmax activation. (a field of use and technological environment MPEP 2106.05(h))
Regarding claim 9,
2A Prong 1: Incorporates the rejection of claim 5.
2A Prong 2: wherein the machine learning model processes the numerical data using a linear activation. (a field of use and technological environment MPEP 2106.05(h))
2B: wherein the machine learning model processes the numerical data using a linear activation. (a field of use and technological environment MPEP 2106.05(h))
Regarding claim 10,
2A Prong 1: Incorporates the rejection of claim 1.
2A Prong 2: wherein at least one of the first segment of the client devices or the second segment of the client devices receives the digital content recommendation. (a field of use and technological environment MPEP 2106.05(h))
2B: wherein at least one of the first segment of the client devices or the second segment of the client devices receives the digital content recommendation. (a field of use and technological environment MPEP 2106.05(h))
Regarding claim 11,
2A Prong 1:
generating a first latent vector representation of a first segment of the client devices and a second latent vector representation of a second segment of the client devices
computing a change vector based on a difference between the first latent vector representation and the second latent vector representation in a latent space of the machine learning model; and (a mathematical concept – calculating a difference between the first vector and the second vector)
combining the change vector with a third latent vector representation in the latent space for a third segment of the client devices; (a mathematical concept – calculating the sum of the difference between the first vector and the second vector and the third vector)
generating an indication of an effect of the second version of the application on a third segment of the client devices based on the combined change vector with the third latent vector representation
2A Prong 2:
A system comprising: a memory component; and a processing device coupled to the memory component, the processing device to perform operations comprising: (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
receiving, via a network, input data describing interactions of client devices included in a group of client devices as part of application version testing of first and second versions of an application; (insignificant extra-solution activity MPEP 2106.05(g)(iii) of gathering statistics)
generating … using an encoder of a machine learning model; (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
generating … using a decoder of the machine learning model; (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
The additional elements as disclosed above alone or in combination do not integrate the judicial exception into practical application as they are mere insignificant extra solution activity, combination of generic computer functions that are restricted to field of use are implemented to perform the disclosed abstract idea above.
2B:
A system comprising: a memory component; and a processing device coupled to the memory component, the processing device to perform operations comprising: (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
receiving, via a network, input data describing interactions of client devices included in a group of client devices as part of application version testing of first and second versions of an application; (indicated as an insignificant extra-solution activity MPEP 2106.05(g)(iii) in Step 2A Prong 2. Therefore, the limitation is re-evaluated in Step 2B as well understood, routine and conventional activity MPEP 2106.05(d)(II)(i) of transmitting or receiving data over a network)
generating … using an encoder of a machine learning model; (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
generating … using a decoder of the machine learning model; (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
The additional elements as disclosed above in combination of the abstract idea are not sufficient to amount to significantly more than the judicial exception as they are well, understood, routine and conventional activity as disclosed in combination of generic computer functions and usage of elements that are restricted to field of use that are implemented to perform the disclosed abstract idea above.
Claim 12 is a system claim having similar limitation to claim 4. Therefore, claim 12 is rejected under the same rationale as of claim 4 above.
Regarding claim 13,
2A Prong 1: wherein the machine learning model is trained using a binary cross-entropy loss for categorical data included in the input data. (a mathematical concept - calculating difference between prediction output and ground truth to update a set of values included in the model)
2A Prong 2: This judicial exception is not integrated into a practical application.
2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
Regarding claim 14,
2A Prong 1: wherein the machine learning model is trained using a mean squared loss for numerical data included in the input data. (a mathematical concept - calculating difference between prediction output and ground truth to update a set of values included in the model)
2A Prong 2: This judicial exception is not integrated into a practical application.
2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
Regarding claim 15,
2A Prong 1: Incorporates the rejection of claim 1.
2A Prong 2: wherein the machine learning model is trained on the input data without an indication of the second version of the application. (mere instructions to apply an exception using a computer MPEP 2106.05(f) – generic training process of a model)
2B: wherein the machine learning model is trained on the input data without an indication of the second version of the application. (mere instructions to apply an exception using a computer MPEP 2106.05(f) – generic training process of a model)
Regarding claim 16,
2A Prong 1:
representing the categorical data and the numerical data as a concatenated vector by batch normalizing the categorical data; (mathematical concept – converting the data to a vector representation by normalizing the data)
generating a first latent vector representation of a first segment of the client devices and a second latent vector representation of a second segment of the client devices based on the concatenated vector
computing a change vector based on a difference between the first latent vector representation and the second latent vector representation in a latent space of the machine learning model; (a mathematical concept – calculating the difference between the first vector and the second vector)
combining the change vector with a third latent vector representation in the latent space for a third segment of the client devices; and (a mathematical concept – calculating the sum of the difference between the first vector and the second vector and the third vector)
generating an indication of an effect of a treatment on a third segment of the client devices by decoding the change vector combined with the third latent vector representation
2A Prong 2:
A non-transitory computer-readable storage medium storing executable instructions, which when executed by a processing device, cause the processing device to perform operations comprising: (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
receiving, via a network, input data describing interactions of client devices included in a group of client devices, the input data includes categorical data and numerical data; (insignificant extra-solution activity MPEP 2106.05(g)(iii) of gathering statistics)
generating using an encoder of a machine learning model (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
generating using a decoder of the machine learning model (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
The additional elements as disclosed above alone or in combination do not integrate the judicial exception into practical application as they are mere insignificant extra solution activity, combination of generic computer functions that are restricted to field of use are implemented to perform the disclosed abstract idea above.
2B:
A non-transitory computer-readable storage medium storing executable instructions, which when executed by a processing device, cause the processing device to perform operations comprising: (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
receiving, via a network, input data describing interactions of client devices included in a group of client devices, the input data includes categorical data and numerical data; (indicated as an insignificant extra-solution activity MPEP 2106.05(g)(iii) in Step 2A Prong 2. Therefore, the limitation is re-evaluated in Step 2B as well understood, routine and conventional activity MPEP 2106.05(d)(II)(i) of transmitting or receiving data over a network)
generating using an encoder of a machine learning model (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
generating using a decoder of the machine learning model (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
The additional elements as disclosed above in combination of the abstract idea are not sufficient to amount to significantly more than the judicial exception as they are well, understood, routine and conventional activity as disclosed in combination of generic computer functions and usage of elements that are restricted to field of use that are implemented to perform the disclosed abstract idea above.
Regarding claim 17,
2A Prong 1: Incorporates the rejection of claim 16.
2A Prong 2: wherein the machine learning model is a variational autoencoder. (a field of use and technological environment MPEP 2106.05(h))
2B: wherein the machine learning model is a variational autoencoder. (a field of use and technological environment MPEP 2106.05(h))
Claim 18 is a non-transitory computer-readable storage medium claim having similar limitation to claim 3. Therefore, claim 18 is rejected under the same rationale as of claim 3 above.
Claim 19 is a non-transitory computer-readable storage medium claim having similar limitation to claim 6. Therefore, claim 19 is rejected under the same rationale as of claim 6 above.
Claim 20 is a non-transitory computer-readable storage medium claim having similar limitation to claim 7. Therefore, claim 18 is rejected under the same rationale as of claim 7 above.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
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.
Claims 1 and 3-4 are rejected under 35 U.S.C. 103 as being unpatentable over Bilello et al. (US 20140257708 A1) in view of VAHDAT et al. (US 20220101144 A1, hereinafter ‘VAHDAT’) and further in view of Wei et al. (US 20190332710 A1, hereinafter ‘Wei’).
Regarding claim 1, Bilello teaches:
receiving, by a processing device via a network, input data describing interactions with digital content by client devices included in a group of client devices, respectively; ([Bilello, 0030-0031], [0078] and [0079] collectively discloses that the biomarkers are collected using devices such as MRI (magnetic resonance imaging), computerized tomography scanning, and magnetic resonance spectroscopy (client devices). The client devices determine and provide scores and test results to the biomarker library. Each x1, x2, x3 … xn are the n parameters corresponds to each medical devices and test results, which also shows which medical device was used to perform the evaluation. [Bilello, 0165] discloses the system extracting parameters from image. The image is the digital content and the system interacts with digital contents by extracting parameters. [Bilello, 0202] discloses receiving patient data and biomarkers from a biomarker library database 710 and a patient database 720. The patient data and biomarkers are processed by a processing engine 730 and stored in output device 740. 710-740 are interpreted as different segment of devices included in a group of client devices)
generating, by the processing device, a first first segment of the client devices and a second second segment of the client devices using [Bilello, 0014] discloses that the algorithm (i.e., the machine learning model) may be a neural network. [Bilello, 0050] discloses that the first representation is ‘a first MDD score generated for a plurality of analytes in a biological sample from the individual, wherein the plurality of analytes comprise one or more HPA axis biomarkers and one or more metabolic biomarkers’ and the second representation is ‘a second MDD score generated for after treatment of the individual for the depression disorder’, and both are generated using the algorithm)
generating, by the processing device, (output) a the third segment of the client devices by [Bilello, 0050] discloses generating a score that indicates whether the treatment was effective on a processing engine and outputting the result on an output device 740 (third segment of the client device) disclosed in [Bilello, 0202]. [Bilello, 0218] and [0219] discloses calculating differences between the groups to determine an effect of a treatment)
However, Bilello does not specifically disclose:
generating, by the processing device, a first latent vector representation of a first segment of the client devices and a second latent vector representation of a second segment of the client devices using an encoder of a machine learning model;
computing, by the processing device, a change vector based on a difference between the first latent vector representation and the second latent vector representation in a latent space of the machine learning model;
combining, by the processing device, the change vector with a third latent vector representation in the latent space for a third segment of the client devices;
generating, by the processing device, digital content recommendation for the third segment of the client devices by decoding the third latent vector representation as combined with the change vector using a decoder of the machine learning model.
Vahdat teaches:
A method comprising:
receiving, by a processing device via a network, input data describing interactions with digital content by [Vahdat, 0029-0030] discloses that the VAEs can be trained on image data and used to produce images, text, music, and other content that can be used on interactive contents such as games, videos, publications, and computer graphics applications)
generating, by the processing device, a first latent vector representation of a first [VAHDAT, 0032] A set of latent variable values sampled from the prior network (includes the second and third latent vector) are collected and inputted into the classifier to generate a ‘reweighting factor’ that captures a difference between the sampled latent variable values and actual latent variable values (i.e., the first latent vector) generated by the encoder network. The reweighting factor (i.e., the change vector) is combined with the sampled latent variable values, and then inputted into the decoder network to generate new “generative output” that is not found in the training dataset)
computing, by the processing device, a change vector based on a difference between the first latent vector representation and the second latent vector representation in a latent space of the machine learning model; and ([VAHDAT, 0032] A set of latent variable values sampled from the prior network (includes the second and third latent vector) are collected and inputted into the classifier to generate a ‘reweighting factor’ that captures a difference between the sampled latent variable values and actual latent variable values (i.e., the first latent vector) generated by the encoder network)
combining, by the processing device, the change vector with a third latent vector representation in the latent space for a third segment of the client devices; ([VAHDAT, 0032] A set of latent variable values sampled from the prior network (includes the second and third latent vector) are collected and inputted into the classifier to generate a ‘reweighting factor’ that captures a difference between the sampled latent variable values and actual latent variable values (i.e., the first latent vector) generated by the encoder network. The reweighting factor (i.e., the change vector) is combined with the sampled latent variable values, and then inputted into the decoder network to generate new “generative output” that is not found in the training dataset)
generating, by the processing device, the third segment of the client by decoding the third latent vector representation as combined with the change vector using a decoder of the machine learning model. ([VAHDAT, 0032] A set of latent variable values sampled from the prior network (includes the second and third latent vector) are collected and inputted into the classifier to generate a ‘reweighting factor’ that captures a difference between the sampled latent variable values and actual latent variable values (i.e., the first latent vector) generated by the encoder network. The reweighting factor (i.e., the change vector) is combined with the sampled latent variable values, and then inputted into the decoder network to generate new “generative output” that is not found in the training dataset)
Before the effective filing date of the invention to a person of ordinary skill in the art, it would have been obvious, having the teachings of Bilello and Vahdat to use the variational autoencoder of Vahdat to implement the prediction method of Bilello. The suggestion and/or motivation for doing so is to improve the accuracy and efficiency of the prediction system by utilizing a machine learning method that specializes in identifying outliers, and comparing latent representation with lower dimensions improves efficiency of the machine learned system.
However, Bilello in view of Vahdat does not specifically disclose:
generating, by the processing device, digital content recommendation for a third segment of the client devices … using the machine learning model.
Wei teaches:
generating, by the processing device, digital content recommendation for a third segment of the client devices … using the machine learning model. ([Wei, 0068 and 0074] collectively disclose a machine learning model that compares two different digital contents to generate the recommendation. [Wei, 0094 and 0106] collectively disclose the system recommending digital contents to a user of a client device for synchronization)
Before the effective filing date of the invention to a person of ordinary skill in the art, it would have been obvious, having the teachings of Bilello, Vahdat, and Wei to use the digital content generation method using a machine learning model of Wei to implement the prediction method of Bilello. The suggestion and/or motivation for doing so is to improve the usability of the prediction system by enabling Bilello’s machine learning based prediction method to provide wider variety of output data (various types of digital content items).
Regarding claim 3, Bilello in view of Vahdat teaches:
The method as described in claim 1, further comprising regularizing the latent space using a Kullback-Leibler divergence loss. ([Vahdat, 0045] The latent space representation q(z|x) is learned by the Kullback-Leibler (KL) divergence)
Regarding claim 4, Bilello in view of Vahdat teaches:
The method as described in claim 1, wherein the machine learning model is a variational autoencoder. ([Vahdat, 0027 and 0032] The trained VAE and classifier are used together to produce generative output that resembles the data in the training dataset)
Claim 2 is rejected under 35 U.S.C. 103 as being unpatentable over Bilello in view of Vahdat in view of Wei and further in view of Li et al. (Li et al, “Learning to Learn from Noisy Labeled Data”, 2019, hereinafter ‘Li’).
Regarding claim 2, Bilello in view of Vahdat and further in view of Wei teaches:
The method as described in claim 1, wherein the machine learning model is trained on the input data [Vahdat, 0029] The VAE is trained on a training dataset that includes tens of thousands to millions of pixels)
However, Bilello in view of Vahdat and further in view of Wei does not specifically disclose:
the machine learning model is trained on the input data without an indication of the digital content recommendation.
Li teaches:
the machine learning model is trained on the input data without an indication of the digital content recommendation. ([Li, page 3, Figure 2 and the explanation] discloses training the machine learning model using synthetic training data without indication (additional label) of whether the training data is modified or not)
Before the effective filing date of the invention to a person of ordinary skill in the art, it would have been obvious, having the teachings of Bilello, Vahdat, Wei and Li to use the method of training the machine learning model on the input data without an indication of the treatment of Li to implement the machine learning based system of Bilello. The suggestion and/or motivation for doing so is to improve the efficiency of the machine learning system by avoiding labeling the input data using treatment indications.
Claims 5-6 and 9 are rejected under 35 U.S.C. 103 as being unpatentable over Bilello in view of Vahdat in view of Wei and further in view of Chaaraoui et al. (US 20210326680 A1, hereinafter ‘Chaaraoui’).
Regarding claim 5, Bilello in view of Vahdat and further in view of Wei teaches:
The method as described in claim 1, wherein the input data includes [Vahdat, 0029] The VAE is trained on a training dataset that includes tens of thousands to millions of pixels. Each pixel includes numbers that represents the color of the pixel, therefore can be interpreted as numerical data)
Chaaraoui teaches:
The method as described in claim 1, wherein the input data includes categorical data and numerical data. ([Chaaraoui, 0052] discloses the training data includes measurement data. [Chaaraoui, 0053] discloses that the input data is derived from test metric, and normalized to numerical data by converting categorical features into numerical features, which implies that the data includes both categorical features and numerical features)
Before the effective filing date of the invention to a person of ordinary skill in the art, it would
have been obvious, having the teachings of Bilello, Vahdat, Wei and Chaaraoui to use the input data describing interactions of client devices and the method of representing categorical data and numerical data as a vector of Chaaraoui to implement the machine learning method of Bilello. The suggestion and/or motivation for doing so is to improve the performance of the effect estimation system by utilizing input data that better represents the interactions between devices and by formatting the input data that can be easily processed by the machine learning model.
Regarding claim 6, Bilello in view of Vahdat in view of Wei and further in view of Chaaraoui teaches:
The method as described in claim 5, wherein the machine learning model is trained using a binary cross-entropy loss for the categorical data. ([Chaaraoui, 0058] discloses that the loss function is a binary cross entropy, and the machine learning model is trained using the loss function. [Chaaraoui, 0053] discloses that the input data is derived from test metric, and normalized to numerical data by converting categorical features into numerical features, which implies that the data includes categorical features)
Regarding claim 9, Bilello in view of Vahdat in view of Wei and further in view of Chaaraoui teaches:
The method as described in claim 5, wherein the machine learning model processes the numerical data using a linear activation. ([Chaaraoui, 0037] discloses processing data using a linear activation function and a batch normalization layer. [Chaaraoui, 0052] discloses the training data includes measurement data. [Chaaraoui, 0053] discloses that the input data is derived from test metric, and normalized to numerical data by converting categorical features into numerical features, which implies that the data includes both categorical features and numerical features)
Claims 7 is rejected under 35 U.S.C. 103 as being unpatentable over Bilello in view of Vahdat in view of Wei in view of Chaaraoui and further in view of Thornton et al. (US 20200218983 A1, hereinafter ‘Thornton’).
Regarding claim 7, Bilello in view of Vahdat in view of Wei and further in view of Chaaraoui teaches:
The method as described in claim 5, wherein the machine learning model is trained [Vahdat, 0029] The VAE is trained on a training dataset that includes tens of thousands to millions of pixels)
However, Bilello in view of Vahdat in view of Wei and further in view of Chaaraoui does not specifically disclose:
wherein the machine learning model is trained using a mean squared loss for the numerical data;
Thornton teaches:
wherein the machine learning model is trained using a mean squared loss for the numerical data; ([Thornton, 0290] discloses that the numerical input vectors are used to train the autoencoder, and a mean squared loss is used as a loss function)
Before the effective filing date of the invention to a person of ordinary skill in the art, it would have been obvious, having the teachings of Bilello, Vahdat, Wei, Chaaraoui, and Thornton to use the method of training the machine learning model using a mean squared loss for the numerical data of Thornton to implement the machine learning method of Bilello. The suggestion and/or motivation for doing so is to improve the efficiency of the machine learning system. The mean-squared loss is easy to implement as the algorithm and the equation is simple.
Claim 8 is rejected under 35 U.S.C. 103 as being unpatentable over Bilello in view of Vahdat in view of Wei in view of Chaaraoui and further in view of Song et al. (US 20220108712 A1, hereinafter ‘Song’).
Regarding claim 8,
Bilello in view of Vahdat in view of Wei and further in view of Chaaraoui teaches:
The method as described in claim 5, wherein the machine learning model processes the categorical data ([Chaaraoui, 0053] discloses that the input data is derived from test metric, and normalized to numerical data by converting categorical features into numerical features, which implies that the data includes both categorical features and numerical features)
Bilello in view of Vahdat in view of Wei and further in view of Chaaraoui does not specifically disclose:
The method as described in claim 5, wherein the machine learning model processes the categorical data using a softmax activation.
Song teaches:
The method as described in claim 5, wherein the machine learning model processes the categorical data using a softmax activation. ([Song, 0042] discloses that the decoder 150 includes a softmax layer set to 2048)
Before the effective filing date of the invention to a person of ordinary skill in the art, it would have been obvious, having the teachings of Bilello, Vahdat, Wei, Chaaraoui, and Song to use the method of processing categorical data using a softmax activation of Song to implement the machine learning method of Bilello. The suggestion and/or motivation for doing so is to implement a recommendation system, as softmax layer is used to determine output data that best fits the input data based on a probability distribution.
Claim 10 is rejected under 35 U.S.C. 103 as being unpatentable over Bilello in view of Vahdat in view of Wei and further in view of Song.
Regarding claim 10, Bilello in view of Vahdat in view of Wei does not specifically disclose:
The method as described in claim 1, wherein at least one of the first segment of the client devices or the second segment of the client devices receives the treatment.
Song teaches:
The method as described in claim 1, wherein at least one of the first segment of the client devices or the second segment of the client devices receives the treatment. ([Song, 0042-0043, 0050 and 0054] discloses generating the output 160 of the system from the WaveNet decoder. The output is a denoised audio representation of the noisy audio. The denoising process is interpreted as the treatment, and the output denoised audio representation is interpreted as the indication of an effect of the denoising, as the treatment can be broadly interpreted as any type of modification or test applied to a device or a human. [Song, 0051] discloses that the loss (i.e., difference between the first latent vector and the second latent vector) is used to update the denoising system including the encoder and the decoder)
Before the effective filing date of the invention to a person of ordinary skill in the art, it would have been obvious, having the teachings of Bilello, Vahdat, Wei, and Song to use the method of receiving treatment using at least one of the first segment of the client devices or the second segment of the client devices of Song to implement the machine learning method of Bilello. The suggestion and/or motivation for doing so is to improve the performance of the treatment recommendation system by training the machine learning model using a comparative data of the untreated devices and treated devices.
Claims 11-12, and 15 are rejected under 35 U.S.C. 103 as being unpatentable over Bilello in view of Vahdat and further in view of Somers et al. (US 20200201322 A1, hereinafter ‘Somers’).
Regarding claim 11, Bilello teaches:
A system comprising: a memory component; and a processing device coupled to the memory component, the processing device to perform operations comprising: ([Bilello, 0202] discloses receiving patient data and biomarkers from a biomarker library database 710 and a patient database 720. The patient data and biomarkers are processed by a processing engine 730 and stored in output device 740. 710-740 are interpreted as different segment of devices included in a group of client devices)
receiving, via a network, input data describing interactions of client devices included in a group of client devices [Bilello, 0030-0031], [0078] and [0079] collectively discloses that the biomarkers are collected using devices such as MRI (magnetic resonance imaging), computerized tomography scanning, and magnetic resonance spectroscopy (client devices). The client devices determine and provide scores and test results to the biomarker library. Each x1, x2, x3 … xn are the n parameters corresponds to each medical devices and test results, which also shows which medical device was used to perform the evaluation (interactions of client devices). [Bilello, 0202] discloses receiving patient data and biomarkers from a biomarker library database 710 and a patient database 720. The patient data and biomarkers are processed by a processing engine 730 and stored in output device 740. 710-740 are interpreted as different segment of devices included in a group of client devices)
generating a first [Bilello, 0014] discloses that the algorithm may be a neural network (i.e., a machine learning model). [Bilello, 0050] discloses that the first representation is ‘a first MDD score generated for a plurality of analytes in a biological sample from the individual, wherein the plurality of analytes comprise one or more HPA axis biomarkers and one or more metabolic biomarkers’ and the second representation is ‘a second MDD score generated for after treatment of the individual for the depression disorder’ and both are generated using the algorithm, which is a neural network machine learning model)
generating an indication of an effect [Bilello, 0050] discloses generating a score that indicates whether the treatment was effective on a processing engine and outputting the result on an output device 740 disclosed in [Bilello, 0202]. [Bilello, 0218] and [0219] discloses calculating differences between the groups to determine an effect of a treatment)
However, Bilello does not specifically disclose:
generating a first latent vector representation of a first segment of the client devices (data) and a second latent vector representation of a second segment of the client devices (data) using an encoder of a machine learning model;
computing a change vector based on a difference between the first latent vector representation and the second latent vector representation in a latent space of the machine learning model;
combining the change vector with a third latent vector representation in the latent space for a third segment of the client devices; and
generating an indication of an effect the second version of the application on a third segment of the client devices based on the change vector using a decoder of the machine learning model.
Vahdat teaches:
receiving, via a network, input data describing interactions of client devices included in a group of client devices[Vahdat, 0029-0030] discloses that the VAEs can be trained on image data and used to produce images, text, music, and other content that can be used on interactive contents such as games, videos, publications, and computer graphics applications)
generating a first latent vector representation of a first [VAHDAT, 0032] A set of latent variable values sampled from the prior network (includes the second and third latent vector) are collected and inputted into the classifier to generate a ‘reweighting factor’ that captures a difference between the sampled latent variable values and actual latent variable values (i.e., the first latent vector) generated by the encoder network. The reweighting factor (i.e., the change vector) is combined with the sampled latent variable values, and then inputted into the decoder network to generate new “generative output” that is not found in the training dataset)
computing a change vector based on a difference between the first latent vector representation and the second latent vector representation in a latent space of the machine learning model; and ([VAHDAT, 0032] A set of latent variable values sampled from the prior network (includes the second and third latent vector) are collected and inputted into the classifier to generate a ‘reweighting factor’ that captures a difference between the sampled latent variable values and actual latent variable values (i.e., the first latent vector) generated by the encoder network)
combining the change vector with a third latent vector representation in the latent space for a third segment of the client devices; and ([VAHDAT, 0032] A set of latent variable values sampled from the prior network (includes the second and third latent vector) are collected and inputted into the classifier to generate a ‘reweighting factor’ that captures a difference between the sampled latent variable values and actual latent variable values (i.e., the first latent vector) generated by the encoder network. The reweighting factor (i.e., the change vector) is combined with the sampled latent variable values, and then inputted into the decoder network to generate new “generative output” that is not found in the training dataset)
generating an indication of an effect of combined change vector with the third latent vector representation using a decoder of the machine learning model. ([VAHDAT, 0032] A set of latent variable values sampled from the prior network (includes the second and third latent vector) are collected and inputted into the classifier to generate a ‘reweighting factor’ that captures a difference between the sampled latent variable values and actual latent variable values (i.e., the first latent vector) generated by the encoder network. The reweighting factor (i.e., the change vector) is combined with the sampled latent variable values, and then inputted into the decoder network to generate new “generative output” that is not found in the training dataset)
Before the effective filing date of the invention to a person of ordinary skill in the art, it would have been obvious, having the teachings of Somers and Vahdat to use the variational autoencoder of Vahdat to implement the effect estimation method of Somers. The suggestion and/or motivation for doing so is to improve the accuracy of the application effect estimation system by utilizing a machine learning method that specializes in identifying outliers.
However, Bilello in view of Vahdat does not specifically disclose:
receiving, via a network, input data describing interactions of client devices included in a group of client devices as part of application version testing of first and second version of an application;
generating an indication of an effect of the second version of the application on a third segment of the client devices.
Somers teaches:
receiving, via a network, input data describing interactions of client devices included in a group of client devices as part of application version testing of first and second version of an application; ([Somers, 0013] discloses that the tag information from data (or data packet) includes a unique identifier which is associated with downstream data as the data propagates from system to system (or subsystem to subsystem). [0027-0028] further supports that the unique identifiers are unique ID for each sensor system. These paragraphs indicates that the tag information describe interactions of client devices (sensor system) as the ID information describes which client device was used to generate and transmit the tag info. [Somers, 0020], [0046], [0086] and [0132] disclose comparing functionality of different software versions (i.e., application versions) by receiving system tag information [0013] and generating system latency graphs 200a and 200b. The comparison may be performed for new software version, new system or subsystem components, or other changes, which indicates that a plurality of client devices may be used)
generating an indication of an effect of the second version of the application on a third segment of the client devices ([Somers, 0020], [0046], [0086] and [0132] disclose comparing functionality of different software versions (i.e., application versions). By generating frequency distribution data (for latency and/or for CPU usage data) for multiple software versions, the system 530 determine how the updated software affects the system)
Before the effective filing date of the invention to a person of ordinary skill in the art, it would have been obvious, having the teachings of Bilello, Vahdat, and Somers to use the application effect estimation method of Somers to implement the prediction method of Bilello. The suggestion and/or motivation for doing so is to improve the usability of the prediction system by enabling Bilello’s machine learning based prediction method to consider wider variety of factors that affects treatment outcomes (such as medical software/application update).
Regarding claim 12, Bilello in view of Vahdat teaches:
The system as described in claim 11, wherein the machine learning model is a variational autoencoder. ([Vahdat, 0027 and 0032] The trained VAE and classifier are used together to produce generative output that resembles the data in the training dataset)
Regarding claim 15, Bilello in view of Vahdat and further in view of Somers teaches:
The system as described in claim 11, wherein the machine learning model is trained on the input data without an indication of the second version of the application. ([Somers, 0099 - 0100] discloses that the output (indication of the second version) is generated based on the input data and learned parameters. [Somers, 0020], [0046], [0086] and [0132] disclose comparing functionality of different software versions (i.e., application versions). By generating frequency distribution data (for latency and/or for CPU usage data) for multiple software versions, the system 530 determine how a new software affects the system. [Somers, 0013] also supports that the generated output (indication of the second version) is new data and generated based on older (not new) data)
Claims 13 is rejected under 35 U.S.C. 103 as being unpatentable over Bilello in view of Vahdat in view of Somers and further in view of Chaaraoui.
Regarding claim 13, Bilello in view of Vahdat and further in view of Somers teaches:
The system as described in claim 11.
However, Bilello in view of Vahdat and further in view of Somers does not specifically disclose:
wherein the machine learning model is trained using a binary cross-entropy loss for categorical data included in the input data.
Chaaraoui teaches:
wherein the machine learning model is trained using a binary cross-entropy loss for categorical data included in the input data. ([Chaaraoui, 0058] discloses that the loss function is a binary cross entropy, and the machine learning model is trained using the loss function. [Chaaraoui, 0053] discloses that the input data is derived from test metric, and normalized to numerical data by converting categorical features into numerical features, which implies that the data includes categorical features)
Before the effective filing date of the invention to a person of ordinary skill in the art, it would
have been obvious, having the teachings of Bilello, Vahdat, Somers, and Chaaraoui to use the input data describing interactions of client devices and the method of representing categorical data and numerical data as a vector of Chaaraoui to implement the application effect estimation method of Bilello. The suggestion and/or motivation for doing so is to improve the performance of the effect estimation system by utilizing input data that better represents the interactions between devices and by formatting the input data that can be easily processed by the machine learning model.
Claim 14 is rejected under 35 U.S.C. 103 as being unpatentable over Bilello in view of Vahdat in view of Somers and further in view of Thornton.
Regarding claim 14, Bilello in view of Vahdat and further in view of Somers teaches:
The system as described in claim 11, wherein the machine learning model is trained [Vahdat, 0029] The VAE is trained on a training dataset that includes tens of thousands to millions of pixels)
However, Bilello in view of Vahdat and further in view of Somers does not specifically disclose:
wherein the machine learning model is trained using a mean squared loss for numerical data included in the input data;
Thornton teaches:
wherein the machine learning model is trained using a mean squared loss for numerical data included in the input data; ([Thornton, 0290] discloses that the numerical input vectors are used to train the autoencoder, and a mean squared loss is used as a loss function)
Before the effective filing date of the invention to a person of ordinary skill in the art, it would have been obvious, having the teachings of Bilello, Vahdat, Somers, and Thornton to use the method of training the machine learning model using a mean squared loss for the numerical data of Thornton to implement the machine learning based system of Bilello. The suggestion and/or motivation for doing so is to improve the efficiency of the machine learning system. The mean-squared loss is easy to implement as the algorithm and the equation is simple.
Claims 16-19 are rejected under 35 U.S.C. 103 as being unpatentable over Bilello in view of Vahdat and further in view of Chaaraoui.
Regarding claim 16, Bilello teaches:
A non-transitory computer-readable storage medium storing executable instructions, which when executed by a processing device, cause the processing device to perform operations comprising: ([Bilello, 0202] discloses receiving patient data and biomarkers from a biomarker library database 710 and a patient database 720. The patient data and biomarkers are processed by a processing engine 730 and stored in output device 740. 710-740 are interpreted as different segment of devices included in a group of client devices)
receiving, via a network, input data describing interactions of client devices included in a group of client devices, the input data includes [Bilello, 0030-0031], [0078] and [0079] collectively discloses that the biomarkers are collected using devices such as MRI (magnetic resonance imaging), computerized tomography scanning, and magnetic resonance spectroscopy (client devices). The client devices determine and provide scores and test results to the biomarker library. Each x1, x2, x3 … xn are the n parameters corresponds to each medical devices and test results, which also shows which medical device was used to perform the evaluation (interactions of client devices). [Bilello, 0202] discloses receiving patient data and biomarkers from a biomarker library database 710 and a patient database 720. The patient data and biomarkers are processed by a processing engine 730 and stored in output device 740. 710-740 are interpreted as different segment of devices included in a group of client devices)
generating a first [Bilello, 0014] discloses that the algorithm may be a neural network. [Bilello, 0050] discloses that the first representation is ‘a first MDD score generated for a plurality of analytes in a biological sample from the individual, wherein the plurality of analytes comprise one or more HPA axis biomarkers and one or more metabolic biomarkers’ and the second representation is ‘a second MDD score generated for after treatment of the individual for the depression disorder’)
generating an indication of an effect of a treatment on the third segment of the client devices by decoding the change vector combined with the third latent vector representation using [Bilello, 0050] discloses generating a score that indicates whether the treatment was effective on a processing engine and outputting the result on an output device 740 disclosed in [Bilello, 0202]. [Bilello, 0218] and [0219] discloses calculating differences between the groups to determine an effect of a treatment)
However, Bilello does not specifically disclose:
the input data includes categorical data and numerical data;
representing the categorical data and the numerical data as a concatenated vector by batch normalizing the categorical data;
generating a first latent vector representation of a first segment of the client devices and a second latent vector representation of a second segment of the client devices based on the concatenated vector using an encoder of a machine learning model;
computing a change vector based on a difference between the first latent vector representation and the second latent vector representation in a latent space of the machine learning model;
combining the change vector with a third latent vector representation in the latent space for a third segment of the client devices; and
generating an indication of an effect of a treatment on a third segment of the client devices based on a difference between the first latent vector representation and the second latent vector representation in a latent space using a decoder of the machine learning model.
Vahdat teaches:
receiving, via a network, input data [Vahdat, 0029] The VAE is trained on a training dataset that includes tens of thousands to millions of pixels. Each pixel includes numbers that represents the color in the pixel, therefore can be interpreted as numerical data)
generating a first latent vector representation of a first segment of the client [VAHDAT, 0032] A set of latent variable values sampled from the prior network (includes the second and third latent vector) are collected and inputted into the classifier to generate a ‘reweighting factor’ that captures a difference between the sampled latent variable values and actual latent variable values (i.e., the first latent vector) generated by the encoder network. The reweighting factor (i.e., the change vector) is combined with the sampled latent variable values, and then inputted into the decoder network to generate new “generative output” that is not found in the training dataset)
computing a change vector based on a difference between the first latent vector representation and the second latent vector representation in a latent space of the machine learning model; ([VAHDAT, 0032] A set of latent variable values sampled from the prior network (includes the second and third latent vector) are collected and inputted into the classifier to generate a ‘reweighting factor’ that captures a difference between the sampled latent variable values and actual latent variable values (i.e., the first latent vector) generated by the encoder network)
combining the change vector with a third latent vector representation in the latent space for a third segment of the client devices; ([VAHDAT, 0032] A set of latent variable values sampled from the prior network (includes the second and third latent vector) are collected and inputted into the classifier to generate a ‘reweighting factor’ that captures a difference between the sampled latent variable values and actual latent variable values (i.e., the first latent vector) generated by the encoder network. The reweighting factor (i.e., the change vector) is combined with the sampled latent variable values, and then inputted into the decoder network to generate new “generative output” that is not found in the training dataset)
generating an indication of an effect of a treatment on the third segment of the client by decoding the change vector combined with the third latent vector representation using a decoder of the machine learning model. ([VAHDAT, 0032] A set of latent variable values sampled from the prior network (includes the second and third latent vector) are collected and inputted into the classifier to generate a ‘reweighting factor’ that captures a difference between the sampled latent variable values and actual latent variable values (i.e., the first latent vector) generated by the encoder network. The reweighting factor (i.e., the change vector) is combined with the sampled latent variable values, and then inputted into the decoder network to generate new “generative output” that is not found in the training dataset)
Before the effective filing date of the invention to a person of ordinary skill in the art, it would have been obvious, having the teachings of Bilello and Vahdat to use the variational autoencoder of Vahdat to implement the treatment effect prediction method of Bilello. The suggestion and/or motivation for doing so is to improve the accuracy and efficiency of the treatment effect estimation system by utilizing a machine learning method that specializes in identifying outliers, and comparing latent representation with lower dimensions improves efficiency of the machine learned system.
However, Bilello in view of Vahdat does not specifically disclose:
input data describing interactions of client devices included in a group of client devices, the input data includes categorical data and numerical data;
representing the categorical data and the numerical data as a concatenated vector by batch normalizing the categorical data;
Chaaraoui teaches:
input data describing interactions [Chaaraoui, 0002, 0013, 0065, 0102; Fig. 1 and Fig. 2] collectively discloses that the input data is used in order to perform the respective mobile network testing which describes telecommunication between many devices, and test or client data is provided to the system. The test data (input data) are collected over time and corresponds to a time-based signal metric obtained by running the test procedures on the at least one testing device that is connected to the mobile network. [Chaaraoui, 0052] discloses the training data includes measurement data (i.e., numerical features). [Chaaraoui, 0053] discloses that the input data is derived from test metric, and normalized to numerical data by converting categorical features into numerical features, which implies that the data includes both categorical features and numerical features)
representing the categorical data and the numerical data as a concatenated vector by batch normalizing the categorical data; ([Chaaraoui, 0037 and 0040] discloses normalizing input data using a batch normalization layer. [Chaaraoui, 0052] discloses the training data includes measurement data (i.e., numerical features). [Chaaraoui, 0053] discloses that the input data is derived from test metric, and normalized to numerical data by converting categorical features into numerical features, which implies that the data includes both categorical features and numerical features)
Before the effective filing date of the invention to a person of ordinary skill in the art, it would
have been obvious, having the teachings of Bilello, Vahdat and Chaaraoui to use the input data describing interactions of client devices and the method of representing categorical data and numerical data as a vector of Chaaraoui to implement the machine learning method of Bilello. The suggestion and/or motivation for doing so is to improve the performance of the effect estimation system by utilizing input data that better represents the interactions between devices and by formatting the input data that can be easily processed by the machine learning model.
Regarding claim 17, Bilello in view of Vahdat teaches:
The non-transitory computer-readable storage medium as described in claim 16, wherein the machine learning model is a variational autoencoder. ([Vahdat, 0027 and 0032] The trained VAE and classifier are used together to produce generative output that resembles the data in the training dataset)
Regarding claim 18, Bilello in view of Vahdat teaches:
The non-transitory computer-readable storage medium as described in claim 16, wherein the latent space is regularized using a Kullback-Leibler divergence loss. ([Vahdat, 0045] The latent space representation q(z|x) is learned by the Kullback-Leibler (KL) divergence)
Regarding claim 19, Bilello in view of Vahdat and further in view of Chaaraoui teaches:
The non-transitory computer-readable storage medium as described in claim 16, wherein the machine learning model is trained using a binary cross- entropy loss for the categorical data. ([Chaaraoui, 0058] discloses that the loss function is a binary cross entropy, and the machine learning model is trained using the loss function. [Chaaraoui, 0053] discloses that the input data is derived from test metric, and normalized to numerical data by converting categorical features into numerical features, which implies that the data includes both categorical features)
Claim 20 is rejected under 35 U.S.C. 103 as being unpatentable over Bilello in view of Vahdat in view of Chaaraoui and further in view of Thornton.
Regarding claim 20, Bilello in view of Vahdat and further in view of Chaaraoui teaches:
The non-transitory computer-readable storage medium as described in claim 16, wherein the machine learning model is trained [Vahdat, 0029] The VAE is trained on a training dataset that includes tens of thousands to millions of pixels)
However, Bilello in view of Vahdat and further in view of Chaaraoui does not specifically disclose:
wherein the machine learning model is trained using a mean squared loss for the numerical data.
Thornton teaches:
wherein the machine learning model is trained using a mean squared loss for the numerical data. ([Thornton, 0290] discloses that the numerical input vectors are used to train the autoencoder, and a mean squared loss is used as a loss function)
Before the effective filing date of the invention to a person of ordinary skill in the art, it would have been obvious, having the teachings of Bilello, Vahdat, Chaaraoui, and Thornton to use the method of training the machine learning model using a mean squared loss for the numerical data of Thornton to implement the machine learning based system of Bilello. The suggestion and/or motivation for doing so is to improve the efficiency of the machine learning system. The mean-squared loss is easy to implement as the algorithm and the equation is simple.
Response to Arguments
Applicant's arguments filed 03/24/2026 have been fully considered but they are not persuasive.
Response to Arguments under 35 U.S.C. 101
Arguments: Applicant asserts that (a) the claims as amended recite specific technical operations that cannot practically be performed in the human mind, (b) the recent Desjardins decision criticized the Board’s approach and examiner’s characterization of the claimed encoder and decoder operation as mere “mental processes” or “mathematical concepts” is the type of overbroad, high-level generality that Desjardins cautions against [page 8-9], (c) the application explicitly identifies technical improvements achieved by the claimed features [0022] and amended to reflect these disclosed improvements [page 10], (d) the application provides detailed technical explanation of how the improvement is achieved e.g., by computing a change vector representing the difference between treated and untreated segments in a regularized latent space, combining the change vector with a third segment’s latent representation, and decoding the combination to estimate treatment effects, (e) In Desjardins, the Panel specifically found that a limitation to adjust the first values of the plurality of parameters to optimize performance of the machine learning model on the second machine learning task while protecting performance of the machine learning model on the first machine learning task” constitutes an improvement to how the machine learning model itself operates, and the claims similarly recites specific technical operations by combining latent vectors and decoding the combination [page 11], (f) the specific ordered combination of encoder operations, latent space vector arithmetic, and decoder operations is not well-understood, routine or conventional activity [page 12].
Examiner’s Response: Examiner respectfully disagrees. Regarding (a), examiner does not contend that ‘receiving, by a processing device via a network’, ‘generating, by the processing device’, ‘computing, by the processing device’, ‘combining, by the processing device’ are mentally performable, but rather that, but for the limitations of ‘generating a first latent vector representation and a second latent representation’ can be practically performed in one’s mind because the claim places no limits on what steps are involved in the ‘generating’ process. The limitation of ‘combining, the change vector with a third latent vector’ is mentally performable as combining data can be performed with the aid of pen and paper. The limitation of ‘generating a digital content recommendation’ also can be performed in one’s mind because recommending a content does not require a computer component.
Regarding (b), Applicant mischaracterizes the nature of Desjardins decision, and Desjardins and the instant application are distinguishable. The Panel founds that Desjardins specifically recites improvements because the limitation "adjust the first values of the plurality of parameters to optimize performance of the machine learning model on the second machine learning task while protecting performance of the machine learning model on the first machine learning task" reflects the improvement disclosed in a paragraph 0021. However, the instant application does not recite such an improvement and the limitations of ‘combining, by the processing device, the change vector with a third latent vector representation in the latent space for a third segment of the client devices;’ merely recites analyzation process to analyze the input data to generate recommendations. The claims contain no limitations directed to improving the machine learning model itself, but rather recite a method for determining contents recommendation and treatment recommendation. As far as Examiner can tell, nothing is done to improve the performance of the machine learning model using these data. Similarly, arguments regarding (e) is not persuasive because the claims contain no limitations directed to improving the ML model.
Regarding (c) and (d), MPEP 2106.05(a) indicates that, “when considering whether additional elements integrate a judicial exception into a practical application, the examiner must evaluate if the disclosure provides sufficient details such that one of ordinary skill in the art would recognize the claimed invention as providing an improvement. If the specification explicitly sets forth an improvement but in a conclusory manner (i.e., a bare assertion of an improvement without the detail necessary to be apparent to a person of ordinary skill in the art), the examiner should not determine the claim improves technology. After the examiner has consulted the specification and determined that the disclosed invention improves technology, the claim must be evaluated to ensure the claim itself reflects the disclosed improvement in technology. Intellectual Ventures I LLC v. Symantec Corp., 838 F.3d 1307, 1316, 120 USPQ2d 1353, 1359 (Fed. Cir. 2016).” As discussed in the Final Rejection mailed on 02/05/2026, the Examiner concluded that the specification [0022] and [0027] set forth the improvement but in a conclusory manner. Even assuming, in arguendo, that the specification discloses an improvement to generative AI, which Examiner does not concede, the claims contain no limitations directed to improving the machine learning model for the similar reasons discussed in arguments (b) and (e).
Regarding (f), the Examiner respectfully disagrees. The claimed elements (1) receiving input data describing client device interactions via a network; (2) generating latent vector representation … using an encoder; and (5) generating a digital content recommendation … using a decoder are well understood, routine, and conventional activity 2106.05(d) in combination of generic computer component MPEP 2106.05(f) to perform the mental process of (2) generating latent vector representation, (4) combining the change vector with a third latent vector representation and (5) generating a digital content recommendation.
Accordingly, the arguments regarding claim 1 are not persuasive. Similarly, the arguments regarding claims 11 and 16 are not persuasive. Therefore, the arguments regarding claims 2-10, 12-15 and 17-20 depend from independent claims 1, 11 and 16 are not persuasive.
Response to Arguments under 35 U.S.C. 103
Applicant’s arguments with respect to claims 1-20 have been considered but are moot because the new ground of rejection does not rely on any reference applied in the prior rejection of record for any teaching or matter specifically challenged in the argument.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JUN KWON whose telephone number is (571)272-2072. The examiner can normally be reached Monday – Friday 7:30AM – 4:30PM ET.
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, Abdullah Kawsar can be reached at (571)270-3169. 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.
/JUN KWON/Examiner, Art Unit 2127
/ABDULLAH AL KAWSAR/Supervisory Patent Examiner, Art Unit 2127
Prosecution Insights