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 .
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claim 9 recites the limitation "the second sample data". There is insufficient antecedent basis for this limitation in the claim. For the purpose of further examination, “second sample data” will be read as “the sample data”
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.
Claim(s) 1-3, 6, 13-15, and 22 are rejected under 35 U.S.C. 103 as being unpatentable over ADACHI (U.S. Pub. No. US 20180181865 A1) in view of KIM (U.S. Pub. No. US 10885387 B1).
Regarding claim 1, ADACHI teaches the invention substantially as claimed, including:
A method for operating a neural network, comprising: setting a random mask layer in a first neural network based on a variable dimension type to obtain a second neural network, ([0068] Random variable xin.sup.DF.sub.j is input data, derives from a transformation of the Bernoulli distribution with dropout in a layer before integrated layer DF, and takes any distribution depending on the structure of the previous layer. Random variable z.sub.j derives from dropout in this integrated layer DF, and is the Bernoulli distribution where z.sub.j=0 with probability p.sub.drop of z.sub.j={0, 1}. These two types of random variables are therefore independent. (obtain a second neural network is intended use)) and training the second neural network by using sample data. ([0006] In the learning phase, the designer prepares training data and causes the neural network to learn neuron weights in the neural network so that a desired specific output is produced from input data having a specific pattern, with use of the training data. (training a network is the same, regardless of if it’s a first, second, or third network.))
While ADACHI does teach setting a random mask layer in a network to get a separate network, it does not explicitly teach:
wherein the random mask layer is used for randomly masking a tensor input to the random mask layer;
However, in analogous art that similarly uses trains networks and masks data, KIM teaches a dropout layer masking tensor input.
wherein the random mask layer is used for randomly masking a tensor input to the random mask layer; ((col 16, lines 36-40) For example, a method of inserting a dropout layer which randomly set at least one element of the input tensor to zero according to a preset rate is commonly applied to the classification network.)
It would have been obvious to a person skilled in the art before the effective filing date of the invention to have combined with KIM’s teaching of masking tensor inputs and, with ADACHI’s teaching of setting a mask layer and training a network, to realize, with a reasonable expectation of success, a method that creates a mask layer and trains a network, as in ADACHI, that uses the mask layer to mask tensor input, as in KIM. A person of ordinary skill would have been motivated to make this combination to be prepared to improve accuracy (KIM col. 1, lines 39-41).
Regarding claim 2, ADACHI further teaches:
The method of claim 1, wherein the variable dimension type comprises one or more of the following types: an input dimension being variable; an output dimension being variable; or an intermediate dimension being variable. ([0068] Random variable xin.sup.DF.sub.j is input data,)
Regarding claim 3, ADACHI further teaches:
The method of claim 2, wherein setting the random mask layer in the first neural network based on the variable dimension type comprises: in case that the variable dimension type is the input dimension being variable, setting a random mask layer before an input layer of the first neural network; or wherein setting the random mask layer in the first neural network based on the variable dimension type comprises: in case that the variable dimension type is the output dimension being variable, setting a random mask layer after an output layer of the first neural network; or wherein setting the random mask layer in the first neural network based on the variable dimension type comprises: in case that the variable dimension type is the intermediate dimension being variable, setting a random mask layer in an intermediate layer of the first neural network. ([0068] Random variable xin.sup.DF.sub.j is input data, derives from a transformation of the Bernoulli distribution with dropout in a layer before integrated layer DF, and takes any distribution depending on the structure of the previous layer.)
Regarding claim 6, ADACHI further teaches:
The method of claim 3, wherein setting the random mask layer in the intermediate layer of the first neural network comprises: in case that only one intermediate dimension is variable, setting a random mask layer before or after an associated intermediate layer; or wherein setting the random mask layer in the intermediate layer of the first neural network comprises: in case that multiple intermediate dimensions are variable, setting a random mask layer before or after each of associated intermediate layers. ([0003] A neural network has a structure in which layers for processing data are arranged. Data is supplied to each layer and subjected to computation set in the layer, and then the processed data is output. In detail, input data from an observed object is first supplied to an input layer of the neural network, processed, and output. The data is then supplied to subsequent layers (intermediate layers) in sequence as input data, processed, and output. [0067] This is the total sum of the list of terms xin.sup.DF.sub.j.sub.Z.sub.jW.sub.i,j (1≤j≤n.sub.Xin.sup.DF) to which bias term b.sub.i is added at the end. Two types of random variables xin.sup.DF.sub.j and z.sub.j are included in this expression, with the rest being fixed values.
[0068] Random variable xin.sup.DF.sub.j is input data, derives from a transformation of the Bernoulli distribution with dropout in a layer before integrated layer DF, and takes any distribution depending on the structure of the previous layer.)
Regarding claim 13, KIM further teaches:
The method of claim 1, wherein randomly masking the tensor input to the random mask layer comprises: determining the tensor input to the random mask layer and a random mask tensor; and randomly masking the tensor input to the random mask layer by using the random mask tensor. (col 16, lines 36-47) For example, a method of inserting a dropout layer which randomly set at least one element of the input tensor to zero according to a preset rate is commonly applied to the classification network. Since the concept of the dropout layer is conventionally adopted to prevent the over-fitting of the classification network during training, the inference after completion of the training is performed without the dropout layer to generate a stable output. On the other hand, in the Monte Carlo method, the inference on the input data is performed multiple times with the dropout layer to generate multiple random outputs and estimate the uncertainty score thereon.)
Regarding claim 14, KIM further teaches:
The method of claim 13, wherein randomly masking the tensor input to the random mask layer is expressed by:X=X
PNG
media_image1.png
87
13
media_image1.png
Greyscale
M, wherein X represents a tensor output from the random mask layer, X represents the tensor input to the random mask layer, 0 represents performing bitwise point multiplication operation on tensors, and M represents the random mask tensor. ((col 16, lines 36-40) For example, a method of inserting a dropout layer which randomly set at least one element of the input tensor to zero according to a preset rate is commonly applied to the classification network. (the expression is simply an explanation of how masking works in board terms, and is thus included in KIM’s mapping))
Regarding claim 15, KIM further teaches:
The method of claim 13, wherein the random mask tensor is a hard-masking tensor; or wherein the random mask tensor is a soft-masking tensor; or wherein the random mask tensor is merged by a hard-masking tensor and a soft- masking tensor.( (col 16, lines 36-47) For example, a method of inserting a dropout layer which randomly set at least one element of the input tensor to zero according to a preset rate is commonly applied to the classification network. Since the concept of the dropout layer is conventionally adopted to prevent the over-fitting of the classification network during training, the inference after completion of the training is performed without the dropout layer to generate a stable output. On the other hand, in the Monte Carlo method, the inference on the input data is performed multiple times with the dropout layer to generate multiple random outputs and estimate the uncertainty score thereon.))
Regarding claim 22, it comprises of limitations similar to those of claim 1 and is therefore rejected for similar rationale.
Claim(s) 8 and 12 are rejected under 35 U.S.C. 103 as being unpatentable over ADACHI (U.S. Pub. No. US 20180181865 A1), KIM (U.S. Pub. No. US 10885387 B1) in further view of SALEH (U.S. Pub. No. US 20220300785 A1).
Regarding claim 8, while ADACHI, as modified by KIM, does teach claim 1, which claim 8 is dependent upon, it does not specifically teach:
The method of claim 1, wherein training the second neural network by using the sample data comprises: in case that dimensions of input data of each sample data are a first threshold, training the second neural network by directly using all of the sample data, wherein the first threshold is a maximum value of dimensions of input data of all the sample data.
However, in analogous art that similarly handles training models, SALEH teaches:
The method of claim 1, wherein training the second neural network by using the sample data comprises: in case that dimensions of input data of each sample data are a first threshold, training the second neural network by directly using all of the sample data, wherein the first threshold is a maximum value of dimensions of input data of all the sample data. ([0049] Similarly, the minimum interaction threshold value may be used to generate training data to be used to train neural network 130 prior to performing the analysis discussed herein. Thus, for example, for a first user 104 (e.g., user 104A) who has had more individual interactions than the minimum interaction threshold value, a first user sequence model for the training date may be generated based on the first X individual interactions, wherein X is equal to the minimum interaction threshold value; )
It would have been obvious to a person skilled in the art before the effective filing date of the invention to have combined with SALEH’s method of training a neural network and, with ADACHI’s, as modified by KIM, teaching of setting a mask layer to mask tensors and training a network, to realize, with a reasonable expectation of success, a method that creates a mask layer and trains a network, as in ADACHI, as modified by KIM, that trains a network using a threshold, as in SALEH. A person of ordinary skill would have been motivated to make this combination to be prepared to improve performance (SALEH [0002]).
Regarding claim 12, SALEH further teaches:
The method of claim 1, wherein training the second neural network by using the sample data comprises: in case that at least one first sample data exists, eliminating the first sample data, and training the second neural network by using all second sample data, wherein the first sample data is sample data whose dimensions of input data are not a first threshold, the second sample data is sample data whose dimensions of input data are the first threshold, and the first threshold is a maximum value of dimensions of input data of all sample data. ([0049] Similarly, the minimum interaction threshold value may be used to generate training data to be used to train neural network 130 prior to performing the analysis discussed herein. Thus, for example, for a first user 104 (e.g., user 104A) who has had more individual interactions than the minimum interaction threshold value, a first user sequence model for the training date may be generated based on the first X individual interactions, wherein X is equal to the minimum interaction threshold value; and for a second user 104 (e.g., user 104B) who has had fewer individual interactions than the minimum interaction threshold value, the second user 104 may be excluded that a training user sequence model is not generated for the second user.)
Claim(s) 9 and 10 are rejected under 35 U.S.C. 103 as being unpatentable over ADACHI (U.S. Pub. No. US 20180181865 A1), KIM (U.S. Pub. No. US 10885387 B1) in further view of GANESAN (U.S. Pub. No. US 20060277234 A1) in further view of SALEH (U.S. Pub. No. US 20220300785 A1).
Regarding claim 9, while ADACHI, as modified by KIM, does teach claim 1, which claim 9 is dependent upon, it does not explicitly teach:
The method of claim 1, wherein training the second neural network by using the sample data comprises: in case that at least one first sample data exists, performing augmentation on the first sample data,
However, in analogous art that similarly handles inputs to a model layer, GANESAN teaches:
The method of claim 1, wherein training the second neural network by using the sample data comprises: in case that at least one first sample data exists, performing augmentation on the first sample data, (([0040] In the event that inputs require greater than 64 bit inputs the embodiment of the processor uses a method, the method comprising: generating a plurality of wide variable shift instructions and a plurality of replace range instructions, padding the most significant input bits with zeros to fit the available input width))
It would have been obvious to a person skilled in the art before the effective filing date of the invention to have combined with GANESAN’s method of data augmentation and, with ADACHI’s, as modified by KIM, teaching of setting a mask layer to mask tensors and training a network, to realize, with a reasonable expectation of success, a method that creates a mask layer and trains a network, as in ADACHI, as modified by KIM, that augments the sample data, as in GANESAN. A person of ordinary skill would have been motivated to make this combination to be prepared to improve performance (GANESAN [0007])
While ADACHI, as modified by KIM and GANESAN, does teach augmenting the sample data, it does not explicitly teach:
and training the second neural network by using all the first sample data and all second sample data, wherein the first sample data is sample data whose dimensions of input data are not a first threshold, the second sample data is sample data whose dimensions of input data are the first threshold, and the first threshold is a maximum value of dimensions of input data of all sample data.
However, in analogous art that similarly trains a model, SALEH teaches:
and training the second neural network by using all the first sample data and all second sample data, wherein the first sample data is sample data whose dimensions of input data are not a first threshold, the second sample data is sample data whose dimensions of input data are the first threshold, and the first threshold is a maximum value of dimensions of input data of all sample data. ([0049] Similarly, the minimum interaction threshold value may be used to generate training data to be used to train neural network 130 prior to performing the analysis discussed herein. Thus, for example, for a first user 104 (e.g., user 104A) who has had more individual interactions than the minimum interaction threshold value, a first user sequence model for the training date may be generated based on the first X individual interactions, wherein X is equal to the minimum interaction threshold value; )
It would have been obvious to a person skilled in the art before the effective filing date of the invention to have combined with SALEH’s method of training a neural network and, with ADACHI’s, as modified by KIM and GANESAN, teaching of setting a mask layer to mask tensors and training a network, to realize, with a reasonable expectation of success, a method that creates a mask layer and trains a network, as in ADACHI, as modified by KIM and GANESAN, that trains a network using a threshold, as in SALEH. A person of ordinary skill would have been motivated to make this combination to be prepared to improve performance (SALEH [0002]).
Regarding claim 10, GANESAN further teaches:
The method of claim 9, wherein performing augmentation on the first sample data comprises: performing high-bit zero padding on input data of the first sample data. ([0040] In the event that inputs require greater than 64 bit inputs the embodiment of the processor uses a method, the method comprising: generating a plurality of wide variable shift instructions and a plurality of replace range instructions, padding the most significant input bits with zeros to fit the available input width)
Claim(s) 11 is rejected under 35 U.S.C. 103 as being unpatentable over ADACHI (U.S. Pub. No. US 20180181865 A1), KIM (U.S. Pub. No. US 10885387 B1) GANESAN (U.S. Pub. No. US 20060277234 A1) SALEH (U.S. Pub. No. US 20220300785 A1) in further view of LI (U.S. Pub. No. US 20140019650 A1)
Regarding claim 11, while ADACHI, as modified by KIM, GANESAN, and SALEH, does teach claim 9, which claim 11 is dependent upon, it does not explicitly teach:
The method of claim 9, wherein performing augmentation on the first sample data comprises: performing low-bit zero padding on input data of the first sample data.
However, in analogous art that similarly handles data augmenting, LI teaches:
The method of claim 9, wherein performing augmentation on the first sample data comprises: performing low-bit zero padding on input data of the first sample data. ([0046] Turning to FIG. 5, a flow diagram 500 depicts a method for buffering and converting data in a multi-write bit-fill FIFO such as that of FIGS. 3 and 4. Following flow diagram 500, a variable length input data block is normalized to the maximum expected length, with data bits aligned to the MSB and with zero-padding as needed at the LSB to achieve uniform length.)
It would have been obvious to a person skilled in the art before the effective filing date of the invention to have combined with LI’s method data augmentation and, with ADACHI’s, as modified by KIM, GANESAN, and SALEH, teaching of setting a mask layer to mask tensors and training a network, to realize, with a reasonable expectation of success, a method that creates a mask layer and trains a network, as in ADACHI, as modified by KIM, GANESAN, and SALEH, that augments the data using padding, as in LI. A person of ordinary skill would have been motivated to make this combination to be prepared to improve data management (LI [0004]).
Claim(s) 18-21 are rejected under 35 U.S.C. 103 as being unpatentable over ADACHI (U.S. Pub. No. US 20180181865 A1), and KIM (U.S. Pub. No. US 10885387 B1) in further view of GOLD (U.S. Pub. No. US 20200082258 A1)
Regarding claim 18, while ADACHI, as modified by KIM, does teach claim 1, which claim 18 is dependent upon, it does not explicitly teach:
The method of claim 1, wherein after training the second neural network by using the sample data, the method further comprises: updating a random mask layer in a trained second neural network with a tensor transformation layer, wherein the tensor transformation layer is used to perform dimension transformation on a tensor input to the tensor transformation layer.
However, in analogous art that similarly handles tensor inputs and layers, GOLD teaches:
The method of claim 1, wherein after training the second neural network by using the sample data, the method further comprises: updating a random mask layer in a trained second neural network with a tensor transformation layer, wherein the tensor transformation layer is used to perform dimension transformation on a tensor input to the tensor transformation layer. ([0072] A tensor-based neural network implementation may be one in which topographic layers of neurons are represented by a set of tensors and the transformations between layers are represented by tensor mathematical operations. Multiple tensors may be used to represent the state of each layer, for example one or more tensors may represent the weights between layers, one tensor can represent the combined (weighted) input (activation) for a layer, and another tensor may represent the output (after a non-linear activation function) for that layer.)
It would have been obvious to a person skilled in the art before the effective filing date of the invention to have combined with GOLD’s method of transforming tensor inputs and, with ADACHI’s, as modified by KIM, teaching of setting a mask layer to mask tensors and training a network, to realize, with a reasonable expectation of success, a method that creates a mask layer and trains a network, as in ADACHI, as modified by KIM, and modifies the tensor inputs of the layers to update each layer, as in GOLD. A person of ordinary skill would have been motivated to make this combination to be improve training performance(GOLD [0004]).
Regarding claim 19, ADACHI further teaches:
The method of claim 18, wherein updating the random mask layer in the trained second neural network with a tensor transformation layer comprises: in case that a random mask layer exists before an input layer of the trained second neural network, (([0068] Random variable xin.sup.DF.sub.j is input data, derives from a transformation of the Bernoulli distribution with dropout in a layer before integrated layer DF, and takes any distribution depending on the structure of the previous layer))
GOLD further teaches:
updating the random mask layer before the input layer of the trained second neural network with a tensor padding layer, wherein the tensor padding layer is used to increase a dimension of a tensor input to the tensor padding layer to a first threshold, and the first threshold is a maximum value of dimensions of input data of all the sample data. ([0171] The output tensors of these different convolutions may then be added 1826 together. The receptive field of the combination of channels which may be produced is illustrated with the linear combination of the corresponding simple feature receptive fields 1828. The illustrated receptive field 1828 may correspond to one like the neuron connection pattern illustrated in FIG. 6A-C. The tensor of the total activation is passed through a threshold operation 1830 to produce a tensor of neuron outputs.)
Regarding claim 20, ADACHI further teaches:
The method of claim 18, wherein updating the random mask layer in the trained second neural network with a tensor transformation layer comprises: in case that a random mask layer exists after an output layer of the trained second neural network, (([0068] Random variable xin.sup.DF.sub.j is input data, derives from a transformation of the Bernoulli distribution with dropout in a layer before integrated layer DF, and takes any distribution depending on the structure of the previous layer. (As seen here, this is not the case. Structuring an either or structure into separate claims does not negate the either or logic of the claims. Here, the case is not fulfilled, and it is shown as not fulfilling by ADACHI, nevertheless, GOLD teaches the rest of the claim.)))
GOLD further teaches:
updating the random mask layer after the output layer of the trained second neural network with a tensor cropping layer, wherein the cropping layer is used to reduce a dimension of a tensor input to the tensor cropping layer to a first target value. ([0078] Any channel or channels may be inverted by a suitable inverse tensor operator (i.e. multiplication by −1, subtraction from the maximum, etc.) and the regular and/or inverted versions may be operated on in separate channels.)
Regarding claim 21, ADACHI further teaches:
The method of claim 18, wherein updating the random mask layer in the trained second neural network with a tensor transformation layer comprises: in case that a random mask layer exists in an intermediate layer of the trained second neural network, (([0068] Random variable xin.sup.DF.sub.j is input data, derives from a transformation of the Bernoulli distribution with dropout in a layer before integrated layer DF, and takes any distribution depending on the structure of the previous layer. (As seen here, this is not the case. Structuring an either or structure into separate claims does not negate the either or logic of the claims. Here, the case is not fulfilled, and it is shown as not fulfilling by ADACHI, nevertheless, GOLD teaches the rest of the claim.)))
GOLD further teaches:
updating the random mask layer in the intermediate layer of the trained second neural network with a cascaded tensor cropping and tensor padding layer, wherein the cascaded tensor cropping and tensor padding layer is used to first reduce a dimension of a tensor input to the cascaded tensor cropping and tensor padding layer to a second target value, and then increase the dimension of the tensor to an original dimension. ([0078] Any channel or channels may be inverted by a suitable inverse tensor operator (i.e. multiplication by −1, subtraction from the maximum, etc.) and the regular and/or inverted versions may be operated on in separate channels. [0171] The output tensors of these different convolutions may then be added 1826 together. The receptive field of the combination of channels which may be produced is illustrated with the linear combination of the corresponding simple feature receptive fields 1828. The illustrated receptive field 1828 may correspond to one like the neuron connection pattern illustrated in FIG. 6A-C. The tensor of the total activation is passed through a threshold operation 1830 to produce a tensor of neuron outputs.)
Claim(s) 43 and 45 are rejected under 35 U.S.C. 103 as being unpatentable over KIM (U.S. Pub. No. US 10885387 B1) in view of GOLD (U.S. Pub. No. US 20200082258 A1)
Regarding claim 43, KIM teaches the invention substantially as claimed, including:
A method for operating a neural network, comprising: obtaining prediction data; and inputting the prediction data to a deployed target neural network to obtain a prediction result output from the target neural network, ((col 16, lines 29-40) Herein, an approach such as Monte Carlo method may be adopted for the second classification module 400 to generate the class score and the uncertainty score. In estimating the uncertainty score of the class prediction, a method of generating randomness on outputs of the classification network to generate multiple random outputs and thus to measure mismatch among the random outputs may exist. There may be various ways to generate the randomness. For example, a method of inserting a dropout layer which randomly set at least one element of the input tensor to zero according to a preset rate is commonly applied to the classification network.))
While KIM does teach obtaining and using prediction data, it does not explicitly teach:
wherein the target neural network comprises at least one tensor transformation layer, and the tensor transformation layer is used to perform dimension transformation on a tensor input to the tensor transformation layer.
However, in analogous art that similarly handles model inputs, GOLD teaches:
wherein the target neural network comprises at least one tensor transformation layer, and the tensor transformation layer is used to perform dimension transformation on a tensor input to the tensor transformation layer. ([0072] A tensor-based neural network implementation may be one in which topographic layers of neurons are represented by a set of tensors and the transformations between layers are represented by tensor mathematical operations. Multiple tensors may be used to represent the state of each layer, for example one or more tensors may represent the weights between layers, one tensor can represent the combined (weighted) input (activation) for a layer, and another tensor may represent the output (after a non-linear activation function) for that layer.)
It would have been obvious to a person skilled in the art before the effective filing date of the invention to have combined with GOLD’s method of transforming tensor inputs and, with KIM’s, teaching obtaining and using prediction data, to realize, with a reasonable expectation of success, a method that obtains and uses prediction data, as in KIM, and modifies the tensor inputs of the layers to update each layer, as in GOLD. A person of ordinary skill would have been motivated to make this combination to improve training performance(GOLD [0004]).
Regarding claim 45, GOLD further teaches:
An electronic device, comprising a memory, a transceiver and a processor, wherein the memory is used to store computer programs, the transceiver is used to transmit and receive data under control of the processor, and the processor is used to read the computer program in the memory and perform the method of claim 43 ([0206] Further, it should be appreciated that modules and/or other appropriate means for performing the methods and techniques described herein can be downloaded and/or otherwise obtained by a user terminal and/or base station as applicable. For example, such a device can be coupled to a server to facilitate the transfer of means for performing the methods described herein. Alternatively, various methods described herein can be provided via storage means (e.g., RAM, ROM, a physical storage medium such as a compact disc (CD) or floppy disk, etc.), such that a user terminal and/or base station can obtain the various methods upon coupling or providing the storage means to the device. Moreover, any other suitable technique for providing the methods and techniques described herein to a device can be utilized.)
Claim(s) 44 are rejected under 35 U.S.C. 103 as being unpatentable over KIM (U.S. Pub. No. US 10885387 B1), GOLD (U.S. Pub. No. US 20200082258 A1) in further view of ADACHI (U.S. Pub. No. US 20180181865 A1)
Regarding claim 44, while KIM, as modified by GOLD, does teach claim 43, which 44 is dependent upon, it does not explicitly teach:
The method of claim 43, wherein the target neural network is the second neural network of claim 1
However, in analogous art that similarly handles models, ADACHI teaches:
The method of claim 43, wherein the target neural network is the second neural network of claim 1. ([0006] In the learning phase, the designer prepares training data and causes the neural network to learn neuron weights in the neural network so that a desired specific output is produced from input data having a specific pattern, with use of the training data.)
It would have been obvious to a person skilled in the art before the effective filing date of the invention to have combined with ADACHI’s model and, with KIM’s, as modified by GOLD, teaching obtaining and using prediction data and transforming tensors, to realize, with a reasonable expectation of success, a method that obtains and uses prediction data and transforms tensors, as in KIM, as modified by GOLD, and trains a model, as in ADACHI. A person of ordinary skill would have been motivated to make this combination to improve training model accuracy (ADACHI [0028]).
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to SKIELER A KOWALIK whose telephone number is (571)272-1850. The examiner can normally be reached 8-5.
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, Mariela D Reyes can be reached at (571)270-1006. 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.
/SKIELER ALEXANDER KOWALIK/Examiner, Art Unit 2142
/HAIMEI JIANG/Primary Examiner, Art Unit 2142
Prosecution Insights