DETAILED ACTION
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 .
Response to Amendment
(Submitted 2/20/2026)
In regard to 103 Rejection
- On Page 6, the applicant argues references based on the amended claims 1 (same as 3 and 4) that recites “ each of which includes a parallel block and a non- parallel block,” and “the non-parallel processing block is a processing block duplicated among the plurality of distributed neural network models and is configured to redundantly execute processing duplicated among the plurality of distributed neural network models” and further argues on Page 7 that the claims are specific structural and functional characteristic of the “non-processing blocks” that overcomes the obviousness. Further, the applicant argues that first amendment with respect to “non-parallel processing block” and “parallel block” that each invidual distributed block itesef comprises both a parallel and” non-parallel block”. The applicant argues in regard to second amendment that the noon-parallel block is both structurally and functionally equivalent instances of these blocks and that duplication indicated signifies that certain parts of the model may not be parallelized or distribute leading to redundant implementations. On Page 8, the applicant argues that the non-parallel processing block is configured to redundantly execute the processing duplicated across the distributed NN causing inefficient resource utilization.
Examiner’s Response
Perhaps as known to a POSITA that a CPU can perform a sequential processing and NPU and GPU (offload from CPU) can perform parallel processing. The examiner submits that GUO teaches distributed neural network in [0020] “ Federated learning is generally a process for training a machine learning model, such as a deep neural network, using decentralized client devices (e.g., mobile devices or other processing nodes) and their local client device-specific datasets without explicitly exchanging client data with a centralized server or other client devices. Advantageously, this enables each client device to retain their data locally, reducing security risk and privacy concerns During federated learning, local models (on the distributed client devices) are trained on local datasets and then training-related parameters (e.g. the weights and biases of a deep neural network) are aggregated by a central server to generate a global model that can then be shared among all the distributed client devices” and in fact, this setup of federated learning which is a type of distributed machine learning where the model is partitioned across devices and trained in a decentralized manner is indeed a distributed NN. However, the examiner submits that the arguments are MOOT as a result of amendments to the claims 1, 3 and 4 and has used new grounds of rejection for consistency to the amended claim limitations. The examiner uses two new references” Shen” and “ALMEIDA”. Reference “ ALMEIDA” teaches redundant execution.
In CONCLUSION, the examiner rejects claims 1-4 under 103 and MOVE the application to FINAL REJECTION.
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 1-4 are rejected under 35 U.S.C. 103 as being unpatentable over
in view of Yichun Shen et al. (hereinafter Shen) US 2022/0180125 A1.
In view of Yuhui GUO et al. (hereinafter GUO) US 2022/0318412 A1,
In view of Mario ALMEIDA et al. (hereinafter ALMEIDA) US 2022/0083386 A1.
In regard to claim 1: (Currently Amended)
Shen discloses:
- A non-transitory computer-readable recording medium storing a machine learning program
of controlling machine learning of a plurality of distributed neural network models generated by dividing a neural network, the machine learning program comprising instructions for causing a processor to execute processing including:
[0006]:
FIG. 3 illustrates an example of a process that, as a result of being performed by a first client, a second client, and a model server, updates a set of model parameters in a distributed machine learning system, in accordance with at least one embodiment;
[0545]:
one or more computer programs or one or more applications) executing collectively on one or more processors, by hardware or combinations thereof. In at least one embodiment, code is stored on a computer-readable storage medium, for example, in form of a computer program comprising a plurality of instructions executable by one or more processors. In at least one embodiment, a computer-readable storage medium is a non-transitory computer-readable storage medium
[0545]:
In at least one embodiment, executable instructions are executed such that different instructions are executed by different processors—for example, a non-transitory computer-readable storage medium store instructions and a main central processing unit (“CPU”) executes some of instructions while a graphics processing unit (“GPU”) executes other instructions. In at least one embodiment, different components of a computer system have separate processors and different processors execute different subsets of instructions.
[0059]:
In at least one embodiment, training of a machine learning system is distributed across a plurality of computer systems, with each computer system having access to a subset of training data.
[0059]:
in at least one embodiment, as model parameters are updated by a computer system that has exclusive permission to do so, updated parameters are distributed to a plurality of computer systems. In at least one embodiment, this allows training data to be retained privately by individual systems of a plurality of computer systems, and not shared with others.
[0070]:
FIG. 3 illustrates an example of a process that, as a result of being performed by a first client computer system, a second client computer system, and a model server computer system, updates a set of model parameters in a distributed machine learning system
[0070]:
In at least one embodiment, model parameters may include coefficients of a functional model, weights of a neural network,
[0089]:
in at least one embodiment, code and/or data storage 901 stores weight parameters and/or input/output data of each layer of a neural network trained or used in conjunction with one or more embodiments during forward propagation of input/output data and/or weight parameters during training and/or inferencing using aspects of one or more embodiments.
[0113]:
In at least one embodiment, data center may use CPUs, application-specific integrated circuits (ASICs), GPUs, FPGAs, or other hardware to perform training and/or inferencing using above-described resources.
[0005]:
FIG. 2 illustrates an example of coordinating learning of a machine-learned model with a round robin scheme, in accordance with at least one embodiment;
PNG
media_image1.png
688
581
media_image1.png
Greyscale
[0062]:
In at least one embodiment, each medical service provider provides a computer system that trains a machine-learned model in turn, and computer systems coordinate training using a token which is passed amongst computer systems to designate which computer system is allowed to update machine-learned model.
(BRI: updating model parameters (including neural network weights, functional model coefficients, or control system parameters) in a distributed system does represent a distributed neural network if the parameters are the weights/biases of a neural network, and the updates are performed across multiple machines in a coordinated way
- an individual noise for [[that]] the each of distributed neural network model to [[a]] the non-parallel processing block in that distributed neural network in [[that]] the each distributed neural network model
[0059]:
In at least one embodiment, training of a machine learning system is distributed across a plurality of computer systems, with each computer system having access to a subset of training data
[0060]:
In at least one embodiment, multiple independent machine-learned models are trained using a same set of data distributed across a plurality of computer systems. In at least one embodiment, a machine-learned model can be divided into multiple sets of independent parameters such that individual sets of independent parameters can be trained independently.
[0169]:
in at least one embodiment, where an image or portion of an image does not include motion, temporal noise reduction performed by video image compositor may use information from a previous image to reduce noise in a current image.
[0169]:
in at least one embodiment, where an image or portion of an image does not include motion, temporal noise reduction performed by video image compositor may use information from a previous image to reduce noise in a current image.
[0175]:
Embodiments described herein allow for multiple neural networks to be performed simultaneously and/or sequentially, and for results to be combined together to enable Level 3-5 autonomous driving functionality
[0155]:
In at least one embodiment, vector processors included in a particular PVA may simultaneously execute different computer vision algorithms, on one image, or even execute different algorithms on sequential images or portions of an image.
[0062]:
in at least one embodiment, model parameters are downloaded to an autonomous vehicle, and a computer system on said autonomous vehicle obtains a lock for model parameters and trains a model using driving data collected by said vehicle.
[0128]:
in at least one embodiment, one or more of camera(s) (e.g., all cameras) may record and provide image data (e.g., video) simultaneously.
[0125]:
In at least one embodiment, techniques described herein may be used to train a machine learning system of an autonomous vehicle by training a model on a vehicle and then uploading an updated model
[0494]:
In at least one embodiment, deployment system 3806 may execute deployment pipelines 3910. In at least one embodiment, deployment pipelines 3910 may include any number of applications that may be sequentially, non-sequentially, or otherwise applied to imaging data (and/or other data types) generated by imaging devices, sequencing devices, genomics devices
(BRI: the image data provided to the model may contain noise- See [0169] and the image data can be processed sequentially [See 0494]).
Shen does not explicitly disclose:
- such that the individual noise for that distributed neural network model is different from the individual noise for other distributed neural network models from among the plurality of distributed neural network models,
- and assigning, to a plurality of processes, the plurality of distributed neural network models added with the individual noise to cause each of the plurality of processes to perform the machine learning on an assigned distributed neural network model from among the plurality of distributed neural network models added with the individual noise.
However, GUO discloses
- such that the individual noise for that distributed neural network model is different from the individual noise for other distributed neural network models from among the plurality of distributed neural network models,
FIG. 1 depicts an example system for distributed machine learning using private variational dropout techniques.
In [0111]:
The depicted components, and others not depicted, may be configured to perform various aspects of the methods described herein.
Example Clauses
In [0112] Clause 1:
A method, comprising: updating a set of parameters of a global machine learning model based on a local data set; pruning the set of parameters based on pruning criteria; computing a noise-augmented set of gradients for a subset of parameters remaining after the pruning, based in part on a noise value; and transmitting the noise-augmented set of gradients to a global model server.
In [0122] Clause 11:
A method comprising: receiving a set of parameters trained using private variational dropout, wherein the private variational dropout comprises: training the set of parameters and a set of noise variances, pruning the set of parameters based on the noise variances, clipping a set of gradients for the set of parameters based on a clipping value, and adding noise to each clipped respective gradient of the set of gradients based on the noise value; instantiating a machine learning model using the set of parameters; and generating an output by processing input data using the instantiated machine learning model.
(BRI: when noise is added to each clipped gradient in a set, each individual noise sample is typically a different)
- and assigning, to a plurality of processes, the plurality of distributed neural network models added with the individual noise to cause each of the plurality of processes to perform the machine learning on an assigned distributed neural network model from among the plurality of distributed neural network models added with the individual noise.
In [0122] Clause 11:
A method comprising: receiving a set of parameters trained using private variational dropout, wherein the private variational dropout comprises: training the set of parameters and a set of noise variances, pruning the set of parameters based on the noise variances, clipping a set of gradients for the set of parameters based on a clipping value, and adding noise to each clipped respective gradient of the set of gradients based on the noise value; instantiating a machine learning model using the set of parameters; and generating an output by processing input data using the instantiated machine learning model.
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Shen and GUO.
Shen teaches distributing the noise to each of the sequential block.
GUO teaches adding noise to each of the plurality of network models.
Shen teaches distributing the noise to each of the sequential block.
One of ordinary skill would have motivation to combine Shen and GUO that can improve processing efficiency (GUO [0024])
Shen and Guo does not explicitly disclose:
- adding, for each distributed neural network model of the plurality of distributed neural network models model each of which includes a parallel block and a non- parallel block,
- the non-parallel processing block is a processing block duplicated among the plurality of distributed neural network models and is configured to redundantly execute processing duplicated among the plurality of distributed neural network models;
However, ALMEIDA discloses:
- adding, for each distributed neural network model of the plurality of distributed neural network models model each of which includes a parallel block and a non- parallel block,
[0013]:
The present techniques may split a network in multiple arbitrary partitions and assign those into different instances to be run. A deep neural network (DNN) may be represented by a Directed Acyclic Graph (DAG), called a dependency or execution graph of a network. Such a graph can be represented as G=(V, E), where V is the set of modules and E represents the data dependencies. Despite the various branches and skip connections in modern DNN architectures, their computation can be run sequentially or in parallel.
[BRI: a data dependence with sequential processing represents a non-parallel block)
[0080]:
Thus, a neural network model may be split and run on a CPU (full precision), GPU (half-precision), NPU (low precision), and/or a DSP (low precision), while at the same time maximising parallel execution. The scheduling step (step S112 in FIG. 1) may comprise scheduling and pipelining the execution of the partitions 402 in a way that each partition takes a similar time to execute, thus maximising throughput.
[0030]:
In some cases, instead of assigning each partition of the neural network to a single computing resource to be executed, the method may comprise assigning each partition of the neural network to be executed by a first computing resource and a second computing resource of the determined subset of computing resources.
[0030]:
in some cases, both the first and second computing resources execute the partition of the neural network in parallel, and the result is taken from whichever computing resource completes the execution first/fastest.
- the non-parallel processing block is a processing block duplicated among the plurality of distributed neural network models and is configured to redundantly execute processing duplicated among the plurality of distributed neural network models;
[0030]:
In some cases, instead of assigning each partition of the neural network to a single computing resource to be executed, the method may comprise assigning each partition of the neural network to be executed by a first computing resource and a second computing resource of the determined subset of computing resources. That is, the method may build-in redundancy into the distributed execution, in case one of the first and second computing resources suddenly is unable to complete the required computation /execution. In this case, when the user device receives, during execution of a partition by the first computing resource and the second resource, a message indicating that the first computing resource is unable to complete the execution, the user device may terminate the execution of the partition by the first resource. This is because the second computing resource can complete the execution
[0062]:
in other cases, the execution of one partition may require another partition to have completed or partly completed. Therefore, whichever way the neural network is divided, the execution of the partitions is scheduled to ensure the neural network can be computed in a time and resource efficient manner.
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Shen , GUO and ALMEIDA.
Shen teaches distributing the noise to each of the sequential block.
GUO teaches adding noise to each of the plurality of network models.
She ALMEIDA teaches a model has parallel and non-parallel block and redundant execution.
One of ordinary skill would have motivation to combine GUO and HOYDIS that may reduce model size and speed up operation using tensor quantization (ALMEIDA [0018])
In regard to claim 2 (Original)
Shen does not explicitly disclose:
- the processing further including: executing dropout processing different for each process, wherein the non-parallel processing block is a dropout layer.
However, GUO discloses:
- the processing further including: executing dropout processing different for each process, wherein the non-parallel processing block is a dropout layer.
In [0012]:
FIG. 2 depicts an example workflow for training machine learning models using private variational dropout techniques.
In [0013]:
FIG. 3 is an example flow diagram illustrating a method for training machine learning models at a client system using private variational dropout.
In [0070] :
The method 300 then proceeds to block 330, where the client system transmits the modified set of gradients to the central server. That is, the client system transmits the pruned subset of gradients, clipped and/or with added noise, to the central server. As discussed above, the central server may aggregate the gradients received from the set of client systems in order to generate an overall set of aggregated updates. These aggregated gradients may then be used to refine the global model. Subsequently, the updated global model may be distributed (e.g., for the next round of training, or for use in runtime).
In [0055]:
Generally, the hyperparameters may include any number and variety of configurable elements affecting the structure of the model and the learning process. For example, for a neural network model, the hyperparameters may include variables such as the learning rate, a dropout rate, and the like. The model structure generally includes the number of layers in the model, the number of elements in each of the layers, the activation function(s) to be used, and the like. In some aspects, the model structure or architecture is specified by the central server, while each client may be allowed to separately select their own training hyperparameters (such as learning rate, dropout rate, and the like
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Shen and GUO.
Shen teaches distributing the noise to each of the sequential block.
GUO teaches adding noise to each of the plurality of network models.
Shen teaches distributing the noise to each of the sequential block.
One of ordinary skill would have motivation to combine Shen and GUO that can improve processing efficiency (GUO [0024])
In regard to claim 3: (Currently Amended)
Shen discloses:
- An information processing apparatus of controlling machine learning of a plurality of distributed neural network models generated by dividing a neural network, the information processing apparatus comprising:
[0089]:
In at least one embodiment, training logic 915 may include, or be coupled to code and/or data storage 901 to store graph code or other software to control timing and/or order, in which weight and/or other parameter information is to be loaded to configure, logic,
[0241]:
In at least one embodiment, processing unit 1630 may be any instruction execution system, apparatus, or device capable of executing instructions.
[0089]:
In at least one embodiment, code, such as graph code, loads weight or other parameter information into processor ALUs based on an architecture of a neural network to which such code corresponds. In at least one embodiment, code and/or data storage 901 stores weight parameters and/or input/output data of each layer of a neural network trained or used in conjunction with one or more embodiments during forward propagation of input/output data and/or weight parameters during training and/or inferencing using aspects of one or more embodiments. In at least one embodiment, any portion of code and/or data storage 901 may be included with other on-chip or off-chip data storage, including a processor's L1, L2, or L3 cache or system memory.
[0005]:
FIG. 2 illustrates an example of coordinating learning of a machine-learned model with a round robin scheme, in accordance with at least one embodiment;
PNG
media_image1.png
688
581
media_image1.png
Greyscale
[0062]:
In at least one embodiment, each medical service provider provides a computer system that trains a machine-learned model in turn, and computer systems coordinate training using a token which is passed amongst computer systems to designate which computer system is allowed to update machine-learned model.
(BRI: updating model parameters (including neural network weights, functional model coefficients, or control system parameters) in a distributed system does represent a distributed neural network if the parameters are the weights/biases of a neural network, and the updates are performed across multiple machines in a coordinated way
- an individual noise for [[that]] the each of distributed neural network model to [[a]] the non-parallel processing block in that distributed neural network in [[that]] the each distributed neural network model
[0059]:
In at least one embodiment, training of a machine learning system is distributed across a plurality of computer systems, with each computer system having access to a subset of training data
[0060]:
In at least one embodiment, multiple independent machine-learned models are trained using a same set of data distributed across a plurality of computer systems. In at least one embodiment, a machine-learned model can be divided into multiple sets of independent parameters such that individual sets of independent parameters can be trained independently.
[0169]:
in at least one embodiment, where an image or portion of an image does not include motion, temporal noise reduction performed by video image compositor may use information from a previous image to reduce noise in a current image.
[0169]:
in at least one embodiment, where an image or portion of an image does not include motion, temporal noise reduction performed by video image compositor may use information from a previous image to reduce noise in a current image.
[0175]:
Embodiments described herein allow for multiple neural networks to be performed simultaneously and/or sequentially, and for results to be combined together to enable Level 3-5 autonomous driving functionality
[0155]:
In at least one embodiment, vector processors included in a particular PVA may simultaneously execute different computer vision algorithms, on one image, or even execute different algorithms on sequential images or portions of an image.
[0062]:
in at least one embodiment, model parameters are downloaded to an autonomous vehicle, and a computer system on said autonomous vehicle obtains a lock for model parameters and trains a model using driving data collected by said vehicle.
[0128]:
in at least one embodiment, one or more of camera(s) (e.g., all cameras) may record and provide image data (e.g., video) simultaneously.
[0125]:
In at least one embodiment, techniques described herein may be used to train a machine learning system of an autonomous vehicle by training a model on a vehicle and then uploading an updated model
[0494]:
In at least one embodiment, deployment system 3806 may execute deployment pipelines 3910. In at least one embodiment, deployment pipelines 3910 may include any number of applications that may be sequentially, non-sequentially, or otherwise applied to imaging data (and/or other data types) generated by imaging devices, sequencing devices, genomics devices
(BRI: the image data provided to the model may contain noise- See [0169] and the image data can be processed sequentially [See 0494]).
Shen does not explicitly disclose:
- such that the individual noise for that distributed neural network model is different from the individual noise for other distributed neural network models from among the plurality of distributed neural network models,
- and assign, to a plurality of processes, the plurality of distributed neural network models added with the individual noise to cause each of the plurality of processes to perform the machine learning on an assigned distributed neural network model from among the plurality of distributed neural network models added with the individual noise.
However, GUO discloses:
- such that the individual noise for that distributed neural network model is different from the individual noise for other distributed neural network models from among the plurality of distributed neural network models,
[0011]:
FIG. 1 depicts an example system for distributed machine learning using private variational dropout techniques.
In [0111]:
The depicted components, and others not depicted, may be configured to perform various aspects of the methods described herein.
Example Clauses
In [0112] Clause 1:
A method, comprising: updating a set of parameters of a global machine learning model based on a local data set; pruning the set of parameters based on pruning criteria; computing a noise-augmented set of gradients for a subset of parameters remaining after the pruning, based in part on a noise value; and transmitting the noise-augmented set of gradients to a global model server.
In [0122] Clause 11:
A method comprising: receiving a set of parameters trained using private variational dropout, wherein the private variational dropout comprises: training the set of parameters and a set of noise variances, pruning the set of parameters based on the noise variances, clipping a set of gradients for the set of parameters based on a clipping value, and adding noise to each clipped respective gradient of the set of gradients based on the noise value; instantiating a machine learning model using the set of parameters; and generating an output by processing input data using the instantiated machine learning model.
(BRI: when noise is added to each clipped gradient in a set, each individual noise sample is typically a different)
- and assign, to a plurality of processes, the plurality of distributed neural network models added with the individual noise to cause each of the plurality of processes to perform the machine learning on an assigned distributed neural network model from among the plurality of distributed neural network models added with the individual noise.
In [0122] Clause 11:
A method comprising: receiving a set of parameters trained using private variational dropout, wherein the private variational dropout comprises: training the set of parameters and a set of noise variances, pruning the set of parameters based on the noise variances, clipping a set of gradients for the set of parameters based on a clipping value, and adding noise to each clipped respective gradient of the set of gradients based on the noise value; instantiating a machine learning model using the set of parameters; and generating an output by processing input data using the instantiated machine learning model.
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Shen and GUO.
Shen teaches distributing the noise to each of the sequential block.
GUO teaches adding noise to each of the plurality of network models.
One of ordinary skill would have motivation to combine Shen and GUO that can improve processing efficiency (GUO [0024])
Shen and Guo do not explicitly disclose:
- add, for each distributed neural network model of the plurality of distributed neural network models model each of which includes a parallel block and a non- parallel block,
- the non-parallel processing block is a processing block duplicated among the plurality of distributed neural network models and is configured to redundantly execute processing duplicated among the plurality of distributed neural network models;
However, ALMEIDA discloses:
- add, for each distributed neural network model of the plurality of distributed neural network models model each of which includes a parallel block and a non- parallel block,
[0013]:
The present techniques may split a network in multiple arbitrary partitions and assign those into different instances to be run. A deep neural network (DNN) may be represented by a Directed Acyclic Graph (DAG), called a dependency or execution graph of a network. Such a graph can be represented as G=(V, E), where V is the set of modules and E represents the data dependencies. Despite the various branches and skip connections in modern DNN architectures, their computation can be run sequentially or in parallel.
(BRI: a data dependence with sequential processing represents a non-parallel block)
[0080]:
Thus, a neural network model may be split and run on a CPU (full precision), GPU (half-precision), NPU (low precision), and/or a DSP (low precision), while at the same time maximising parallel execution. The scheduling step (step S112 in FIG. 1) may comprise scheduling and pipelining the execution of the partitions 402 in a way that each partition takes a similar time to execute, thus maximising throughput.
[0030]:
In some cases, instead of assigning each partition of the neural network to a single computing resource to be executed, the method may comprise assigning each partition of the neural network to be executed by a first computing resource and a second computing resource of the determined subset of computing resources.
[0030]:
in some cases, both the first and second computing resources execute the partition of the neural network in parallel, and the result is taken from whichever computing resource completes the execution first/fastest.
- the non-parallel processing block is a processing block duplicated among the plurality of distributed neural network models and is configured to redundantly execute processing duplicated among the plurality of distributed neural network models;
[0030]:
In some cases, instead of assigning each partition of the neural network to a single computing resource to be executed, the method may comprise assigning each partition of the neural network to be executed by a first computing resource and a second computing resource of the determined subset of computing resources. That is, the method may build-in redundancy into the distributed execution, in case one of the first and second computing resources suddenly is unable to complete the required computation /execution.In this case, when the user device receives, during execution of a partition by the first computing resource and the second resource, a message indicating that the first computing resource is unable to complete the execution, the user device may terminate the execution of the partition by the first resource. This is because the second computing resource can complete the execution
[0062]:
in other cases, the execution of one partition may require another partition to have completed or partly completed. Therefore, whichever way the neural network is divided, the execution of the partitions is scheduled to ensure the neural network can be computed in a time and resource efficient manner.
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Shen, GUO and ALMEIDA.
Shen teaches distributing the noise to each of the sequential block.
GUO teaches adding noise to each of the plurality of network models.
ALMEIDA teaches a model has parallel and non-parallel block and redundant execution.
One of ordinary skill would have motivation to combine Shen, GUO and ALMEIIDA that may reduce model size and speed up operation using tensor quantization (ALMEIDA [0018])
In regard to claim 4: (Currently Amended)
Shen discloses:
- A computer-implemented method of controlling machine learning of a plurality of distributed neural network models generated by dividing a neural network, the machine learning program comprising instructions for causing a processor to execute processing including:
[0005]:
FIG. 2 illustrates an example of coordinating learning of a machine-learned model with a round robin scheme, in accordance with at least one embodiment;
PNG
media_image1.png
688
581
media_image1.png
Greyscale
[0062]:
In at least one embodiment, each medical service provider provides a computer system that trains a machine-learned model in turn, and computer systems coordinate training using a token which is passed amongst computer systems to designate which computer system is allowed to update machine-learned model.
(BRI: updating model parameters (including neural network weights, functional model coefficients, or control system parameters) in a distributed system does represent a distributed neural network if the parameters are the weights/biases of a neural network, and the updates are performed across multiple machines in a coordinated way
Shen does not explicitly disclose:
- such that the individual noise for that distributed neural network model is different from the individual noise for other distributed neural network models from among the plurality of distributed neural network models,
- and assigning, to a plurality of processes, the plurality of distributed neural network models added with the individual noise to cause each of the plurality of processes to perform the machine learning on an assigned distributed neural network model from among the plurality of distributed neural network models added with the individual noise.
However, GUO discloses:
- such that the individual noise for that distributed neural network model is different from the individual noise for other distributed neural network models from among the plurality of distributed neural network models,
[0011]:
FIG. 1 depicts an example system for distributed machine learning using private variational dropout techniques.
In [0111]:
The depicted components, and others not depicted, may be configured to perform various aspects of the methods described herein.
Example Clauses
In [0112] Clause 1:
A method, comprising: updating a set of parameters of a global machine learning model based on a local data set; pruning the set of parameters based on pruning criteria; computing a noise-augmented set of gradients for a subset of parameters remaining after the pruning, based in part on a noise value; and transmitting the noise-augmented set of gradients to a global model server.
In [0122] Clause 11:
A method comprising: receiving a set of parameters trained using private variational dropout, wherein the private variational dropout comprises: training the set of parameters and a set of noise variances, pruning the set of parameters based on the noise variances, clipping a set of gradients for the set of parameters based on a clipping value, and adding noise to each clipped respective gradient of the set of gradients based on the noise value; instantiating a machine learning model using the set of parameters; and generating an output by processing input data using the instantiated machine learning model.
(BRI: when noise is added to each clipped gradient in a set, each individual noise sample is typically a different)
- and assigning, to a plurality of processes, the plurality of distributed neural network models added with the individual noise to cause each of the plurality of processes to perform the machine learning on an assigned distributed neural network model from among the plurality of distributed neural network models added with the individual noise.
In [0122] Clause 11:
A method comprising: receiving a set of parameters trained using private variational dropout, wherein the private variational dropout comprises: training the set of parameters and a set of noise variances, pruning the set of parameters based on the noise variances, clipping a set of gradients for the set of parameters based on a clipping value, and adding noise to each clipped respective gradient of the set of gradients based on the noise value; instantiating a machine learning model using the set of parameters; and generating an output by processing input data using the instantiated machine learning model.
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Shen and GUO.
Shen teaches distributing the noise to each of the sequential block.
GUO teaches adding noise to each of the plurality of network models.
One of ordinary skill would have motivation to combine Shen and GUO that can improve processing efficiency (GUO [0024])
Shen and GUO do not explicitly disclose:
- adding, for each distributed neural network model of the plurality of distributed neural network models model each of which includes a parallel block and a non- parallel block,
- the non-parallel processing block is a processing block duplicated among the plurality of distributed neural network models and is configured to redundantly execute processing duplicated among the plurality of distributed neural network models;
However, ALMEIDA discloses:
- adding, for each distributed neural network model of the plurality of distributed neural network models model each of which includes a parallel block and a non- parallel block,
[0013]:
The present techniques may split a network in multiple arbitrary partitions and assign those into different instances to be run. A deep neural network (DNN) may be represented by a Directed Acyclic Graph (DAG), called a dependency or execution graph of a network. Such a graph can be represented as G=(V, E), where V is the set of modules and E represents the data dependencies. Despite the various branches and skip connections in modern DNN architectures, their computation can be run sequentially or in parallel.
(BRI: a data dependent sequential processing is a non-parallel processing)
[0080]:
Thus, a neural network model may be split and run on a CPU (full precision), GPU (half-precision), NPU (low precision), and/or a DSP (low precision), while at the same time maximising parallel execution. The scheduling step (step S112 in FIG. 1) may comprise scheduling and pipelining the execution of the partitions 402 in a way that each partition takes a similar time to execute, thus maximising throughput.
[0030]:
in some cases, both the first and second computing resources execute the partition of the neural network in parallel, and the result is taken from whichever computing resource completes the execution first/fastest.
- the non-parallel processing block is a processing block duplicated among the plurality of distributed neural network models and is configured to redundantly execute processing duplicated among the plurality of distributed neural network models;
[0030]:
In some cases, instead of assigning each partition of the neural network to a single computing resource to be executed, the method may comprise assigning each partition of the neural network to be executed by a first computing resource and a second computing resource of the determined subset of computing resources. That is, the method may build-in redundancy into the distributed execution, in case one of the first and second computing resources suddenly is unable to complete the required computation /execution. In this case, when the user device receives, during execution of a partition by the first computing resource and the second resource, a message indicating that the first computing resource is unable to complete the execution, the user device may terminate the execution of the partition by the first resource. This is because the second computing resource can complete the execution
[0062]:
in other cases, the execution of one partition may require another partition to have completed or partly completed. Therefore, whichever way the neural network is divided, the execution of the partitions is scheduled to ensure the neural network can be computed in a time and resource efficient manner.
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Shen, GUO and ALMEIDA.
Shen teaches distributing the noise to each of the sequential block.
GUO teaches adding noise to each of the plurality of network models.
ALMEIDA teaches a model has parallel and non-parallel block and redundant execution.
One of ordinary skill would have motivation to combine Shen, GUO and ALMEIIDA that may reduce model size and speed up operation using tensor quantization (ALMEIDA [0018])
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Shen, GUO and ALMEIDA.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the
examiner should be directed to TIRUMALE KRISHNASWAMY RAMESH whose telephone number is (571)272-4605. The examiner can normally be reached by phone.
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, Li B Zhen can be reached on phone (571-272-3768). 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.
/TIRUMALE K RAMESH/Examiner, Art Unit 2121
/Li B. Zhen/Supervisory Patent Examiner, Art Unit 2121
Prosecution Insights