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 .
Specification
The disclosure is objected to because of the following informalities:
On page 9, paragraphs 2 and 3 are nearly identical.
On Page 9 paragraph 3 line 22, reference number 248 is referred to as “network control process” when 248 was previously referred to as “machine learning process” in both the specification and drawings
On page 10 paragraph 1 line 10, the word "to" should be added in between "the ratio of true positives [to] the sum of true and false positives".
Appropriate correction is required.
Information Disclosure Statement
The information disclosure statement (IDS) submitted on 02/28/2024 is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claims 1, 4-5, 7-8, 11, 14-15, 17-18, and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Li et al (US Pat. Pub. 2024/0311193; hereinafter referred to as Li) in view of Bajaj (US Pat. Pub. 2019/0036828).
As per claim 1, Li teaches a method comprising: identifying, by the controller (e.g. Li: [0179] when executed by the processor, implements a method for allocating a computing task of a neural network in heterogeneous resources. Please note that the processor is one possible embodiment of the Applicant’s controller.) performance characteristics of different processing paths across the partitioned neural network (e.g. Li: [0006] discloses acquiring task information of the computing task of the neural network and resource information of the heterogeneous resources configured for executing the computing task. [0011] discloses the resource information includes a running speed of each resource among the heterogeneous resources. Please note running speed and heterogeneous resources corresponds to the Applicant’s performance characteristics and processing paths, respectively.); selecting, by the controller, a particular path from among the different processing paths to process the task request, based on(e.g. Li: [0010] discloses selecting a target allocation path according to the value of the loss function corresponding to each allocation path. [0029] discloses obtaining a value of a loss function corresponding to each allocation path according to the task processing cost corresponding to each subtask in each allocation path.); and sending, by controller, the task request for processing by the partitioned neural network along the particular path (e.g. Li: [0156] sending the target allocation path to the scheduling server, whereby the scheduling server executes task scheduling according to the target allocation path).
Li does not teach determining, by a controller, a level of service for a task request for processing by a partitioned neural network executed across a plurality of devices. Nor does Li teach selecting a particular path based on the level of service.
However, Bajaj does teach determining a level of service for a task request (e.g. Bajaj: [0004] discloses the method may additionally include identifying a set of performance thresholds associated with the classification of the application. [0014] discloses an SLA may include an agreed upon threshold level for one or more network performance metrics, such as bandwidth, availability, jitter, latency, loss, and/or others. [0023] discloses a table showing that SLA and level of service are the same because they consider the same factors.); and selecting a particular path based on the level of service (e.g. Bajaj: [0021] A policy may include a rule or set of rules bearing on the handling of network traffic, such as routing, priority, media, etc. In some embodiments, the policies may include SLAs for various data flows. [0014] discloses in at least some embodiments, a tunnel for an application may be changed based on an SLA.).
Li and Bajaj are in the same field of endeavor in terms of path selection and therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the teachings of Li to incorporate the teachings of Bajaj to start determining a level of service for a task request and selecting a particular path from among the different processing paths to process the task request, based on the level of service. This would allow the network to consider factors beyond just speed and availability when deciding which path the request gets allocated to. This improves the flexibility of path selection by letting users give more specific input on what task should be assigned where.
As per claim 4, Li-Bajaj teaches claim 1 as applied above. Bajaj further teaches wherein the level of service is based on a type of request associated with the task request or a data type associated with the task request (e.g. Baja: Please note [0023] discloses a table which gives classifications to different applications. So, applications that deal with audio data would get a Real-time classification while streaming video would get a Streaming classification. These classifications then affect the SLA’s.).
Li and Bajaj are in the same field of endeavor in terms of path selection and therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the teachings of Li to incorporate the teachings of Bajaj to wherein the level of service is based on a type of request associated with the task request or a data type associated with the task request. This improves the flexibility of path selection by letting users give more specific input on what task should be assigned where. In this case, more resource-intensive data types like video and audio streaming would be given a higher quality of service.
As per claim 5, Li-Bajaj teaches claim 1 as applied above. Bajaj teaches wherein the level of service is based on a requestor that issued the task request (e.g. Bajaj: Please note [0023] discloses a table which gives classifications to different applications. In addition to data type, classifications could be based on the identity of the application. So, data from MS Office gets classified as Mission Critical Data while Social Forums gets classified as Scavenger. These classifications then affect the SLA’s.).
Li and Bajaj are in the same field of endeavor in terms of path selection and therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the teachings of Li to incorporate the teachings of Bajaj to wherein the level of service is based on a requestor that issued the task request. This improves the flexibility of path selection by letting users give more specific input on what task should be assigned where. In this case, the network would be able to figure out the importance level of the requestor and then choose the amount and quality of resources to be given.
As per claim 7, Li-Bajaj teaches claim 1 as applied above. Li further teaches the performance characteristics are indicative of at least one of: path latencies of the different processing paths, latency metrics for the plurality of devices (e.g. Li: [0083] discloses determining the execution cost corresponding to each allocation mode according to the running speed of each resource. Please note that running speed corresponds to the Applicant’s latency.), or queue information for the plurality of devices (e.g. Li: [0084] discloses determining, according to the task execution sequence, a layer of the neural network to which the resource allocated to each subtask belongs. Please note the task execution sequence is a queue.).
As per claim 8, Li-Bajaj teaches claim 1 and claim 7 as applied above. Bajaj further teaches applying, a label to the task request indicative of the level of service (e.g. Bajaj: Please note that [0023] discloses a table giving classifications to different tasks and requests.).
Li and Bajaj are in the same field of endeavor in terms of path selection and therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the teachings of Li to incorporate the teachings of Bajaj to make it so that sending the task request for processing along the particular path comprises: applying, a label to the task request indicative of the level of service. Labels can be used to determine the level of service the tasks deserve. This could speed up the decision-making process by quickly figuring out whether the task is important or not.
As per claim 11, Li teaches an apparatus, comprising: a network interface to communicate with a computer network; a processor coupled to the network interface and configured to execute one or more processes; and a memory configured to store a process that is executed by the processor (e.g. Li: [0197] teaches a computer device that includes a processor, a memory, a network interface and a database through which are connected through a system bus.), the process when executed configured to: identify performance characteristics of different processing paths across the partitioned neural network (e.g. Li: [0034] discloses an acquisition module, configured for acquiring task information of the computing task of the neural network and resource information of the heterogeneous resources configured for executing the computing task. [0011] discloses the resource information includes a running speed of each resource among the heterogeneous resources. Please note running speed and heterogeneous resources corresponds to the Applicant’s performance characteristics and processing paths, respectively.); select, based on (e.g. Li: [0010] discloses selecting a target allocation path according to the value of the loss function corresponding to each allocation path. [0029] discloses obtaining a value of a loss function corresponding to each allocation path according to the task processing cost corresponding to each subtask in each allocation path.); and send the task request for processing by the partitioned neural network along the particular path (e.g. Li: [0156] discloses sending the target allocation path to the scheduling server, whereby the scheduling server executes the task scheduling according to the target allocation path.)
Li does not teach a processor when executed configured to determine a level of service for a task request for processing by a partitioned neural network executed across a plurality of devices. Nor does Li teach based on the level of service, selecting a particular path.
However, Bajaj does teach to determine a level of service for a task request (e.g. Bajaj: [0014] discloses an SLA may include an agreed upon threshold level for one or more network performance metrics, such as bandwidth, availability, jitter, latency, loss, and/or others. [0023] discloses a table showing that SLA and level of service are the same because they consider the same factors.); and select, based on the level of service for the task request, a particular path from among the different processing paths to process the task request (e.g. Bajaj: [0021] A policy may include a rule or set of rules bearing on the handling of network traffic, such as routing, priority, media, etc. In some embodiments, the policies may include SLAs for various data flows. [0014] discloses in at least some embodiments, a tunnel for an application may be changed based on an SLA.).
Li and Bajaj are in the same field of endeavor in terms of path selection and therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the teachings of Li to incorporate the teachings of Bajaj to determine a level of service for a task request and select, based on the level of service for the task request, a particular path from among the different processing paths to process the task request. This would allow the network to consider factors beyond just speed and availability when deciding which path the request gets allocated to. This improves the flexibility of path selection by letting users give more specific input on what task should be assigned where.
As per claim 14, Li-Bajaj teaches claim 11 as applied above. Bajaj further teaches wherein the level of service is based on a type of request associated with the task request or a data type associated with the task request (e.g. Baja: Please note [0023] discloses a table which gives classifications to different applications. So, applications that deal with audio data would get a Real-time classification while streaming video would get a Streaming classification. These classifications then affect the SLA’s.).
Li and Bajaj are in the same field of endeavor in terms of path selection and therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the teachings of Li to incorporate the teachings of Bajaj to wherein the level of service is based on a type of request associated with the task request or a data type associated with the task request. This improves the flexibility of path selection by letting users give more specific input on what task should be assigned where. In this case, more resource-intensive data types like video and audio streaming would be given a higher quality of service.
As per claim 15, Li-Bajaj teaches claim 11 as applied above. Bajaj teaches wherein the level of service is based on a requestor that issued the task request (e.g. Bajaj: Please note [0023] discloses a table which gives classifications to different applications. In addition to data type, classifications could be based on the identity of the application. So, data from MS Office gets classified as Mission Critical Data while Social Forums gets classified as Scavenger. These classifications then affect the SLA’s.).
Li and Bajaj are in the same field of endeavor in terms of path selection and therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the teachings of Li to incorporate the teachings of Bajaj to wherein the level of service is based on a requestor that issued the task request. This improves the flexibility of path selection by letting users give more specific input on what task should be assigned where. In this case, the network would be able to figure out the importance level of the requestor and then choose the amount and quality of resources to be given.
As per claim 17, Li-Bajaj teaches claim 11 as applied above. Li further teaches the performance characteristics are indicative of at least one of: path latencies of the different processing paths, latency metrics for the plurality of devices (e.g. Li: [0083] discloses determining the execution cost corresponding to each allocation mode according to the running speed of each resource. Please note that running speed corresponds to the Applicant’s latency.), or queue information for the plurality of devices (e.g. Li: [0084] discloses determining, according to the task execution sequence, a layer of the neural network to which the resource allocated to each subtask belongs. Please note the task execution sequence is a queue.).
As per claim 18, Li-Bajaj teaches claim 11 and claim 17 as applied above. Bajaj further teaches applying, a label to the task request indicative of the level of service (e.g. Bajaj: Please note that [0023] discloses a table giving classifications to different tasks and requests.).
Li and Bajaj are in the same field of endeavor in terms of path selection and therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the teachings of Li to incorporate the teachings of Bajaj to make it so that sending the task request for processing along the particular path comprises: applying, a label to the task request indicative of the level of service. Labels can be used to determine the level of service the tasks deserve. This could speed up the decision-making process by quickly figuring out whether the task is important or not.
As per claim 20, Li teaches a tangible, non-transitory, computer-readable medium storing program instructions that cause a controller (e.g. Li: [0179] when executed by the processor, implements a method for allocating a computing task of a neural network in heterogeneous resources. Please note that the processor is one possible embodiment of the Applicant’s controller.) to execute a process (e.g. Li: [0040] discloses one or more non-volatile computer-readable storage media configured for storing a computer program executable by a processor). comprising: identifying, by the controller performance characteristics of different processing paths across the partitioned neural network (e.g. Li: [0006] discloses acquiring task information of the computing task of the neural network and resource information of the heterogeneous resources configured for executing the computing task. [0011] discloses the resource information includes a running speed of each resource among the heterogeneous resources. Please note running speed and heterogeneous resources corresponds to the Applicant’s performance characteristics and processing paths, respectively.); selecting, by the controller, a particular path from among the different processing paths to process the task request, based onpaths (e.g. Li: [0010] discloses selecting a target allocation path according to the value of the loss function corresponding to each allocation path. [0029] discloses obtaining a value of a loss function corresponding to each allocation path according to the task processing cost corresponding to each subtask in each allocation path.); and sending, by controller, the task request for processing by the partitioned neural network along the particular path (e.g. Li: [0156] sending the target allocation path to the scheduling server, whereby the scheduling server executes task scheduling according to the target allocation path).
Li does not teach determining, by a controller, a level of service for a task request for processing by a partitioned neural network executed across a plurality of devices. Nor does Li teach selecting a particular path based on the level of service.
However, Bajaj does teach determining a level of service for a task request (e.g. Bajaj: [0004] discloses the method may additionally include identifying a set of performance thresholds associated with the classification of the application. [0014] discloses an SLA may include an agreed upon threshold level for one or more network performance metrics, such as bandwidth, availability, jitter, latency, loss, and/or others. [0023] discloses a table showing that SLA and level of service are the same because they consider the same factors.); and selecting a particular path based on the level of service (e.g. Bajaj: [0021] A policy may include a rule or set of rules bearing on the handling of network traffic, such as routing, priority, media, etc. In some embodiments, the policies may include SLAs for various data flows. [0014] discloses in at least some embodiments, a tunnel for an application may be changed based on an SLA.).
Li and Bajaj are in the same field of endeavor in terms of path selection and therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the teachings of Li to incorporate the teachings of Bajaj to start determining a level of service for a task request and selecting a particular path from among the different processing paths to process the task request, based on the level of service. This would allow the network to consider factors beyond just speed and availability when deciding which path the request gets allocated to. This improves the flexibility of path selection by letting users give more specific input on what task should be assigned where.
Claims 9 and 19 are rejected under 35 U.S.C. 103 as being unpatentable over Li-Bajaj in view of Chatterjee et al (US Pat. Pub. 20240069964; priority date Aug. 24, 2022; hereinafter referred to as Chatterjee).
As per claim 9, Li-Bajaj teaches claim 1 as applied above. But they do not teach wherein a device from among the plurality of devices prioritizes the task request for processing based on the label.
However, Chatterjee does teach wherein a device from among the plurality of devices prioritizes the task request for processing based on the label (e.g. Chatterjee: [0060] discloses to schedule jobs using said labels to select nodes that match the job size and have the lowest latency available. For example, if a job requires two GPUs, the scheduler will schedule it on the first node that is labeled to run two GPU jobs, has 2 GPUs available, and has the lowest latency. Please note having a scheduler select a specific node to perform a job corresponds to Applicant’s a plurality of device prioritizing a task.).
Li-Bajaj and Chatterjee are in the same field of endeavor in terms of work distribution and therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the teachings of Li-Bajaj to incorporate the teachings of Chatterjee to wherein a device from among the plurality of devices prioritizes the task request for processing based on the label. This speeds up the processing time for important tasks because they are prioritized thanks to their labels.
As per claim 19, Li-Bajaj teaches claim 11 as applied above. But they do not teach wherein a device from among the plurality of devices prioritizes the task request for processing based on the label.
However, Chatterjee does teach wherein a device from among the plurality of devices prioritizes the task request for processing based on the label (e.g. Chatterjee: [0060] discloses to schedule jobs using said labels to select nodes that match the job size and have the lowest latency available. For example, if a job requires two GPUs, the scheduler will schedule it on the first node that is labeled to run two GPU jobs, has 2 GPUs available, and has the lowest latency. Please note having a scheduler select a specific node to perform a job corresponds to Applicant’s a plurality of device prioritizing a task.).
Li-Bajaj and Chatterjee are in the same field of endeavor in terms of work distribution and therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the teachings of Li-Bajaj to incorporate the teachings of Chatterjee to wherein a device from among the plurality of devices prioritizes the task request for processing based on the label. This speeds up the processing time for important tasks because they are prioritized thanks to their labels.
Claims 2 and 12 are rejected under 35 U.S.C. 103 as being unpatentable over Li-Bajaj in view of Fargo et al (US Pat. Pub. 20190243953; hereinafter referred to as Fargo).
As per claim 2 Li-Bajaj teaches claim 1 as applied above but does not explicitly teach wherein a first device in the plurality of devices executes a partition of the partitioned neural network and a second device in the plurality of devices executes a copy of that partition.
However, Fargo teaches wherein a first device in the plurality of devices executes a partition and a second device in the plurality of devices executes a copy of that partition (e.g. Fargo: [0024] discloses after a workload is submitted by controller 122 to two or more of nodes 130-0 to 130-n, nodes will perform the workload and provide results that are accessible to controller 122. Controller 122 can collect the results and apply a voting mechanism technique to identify any anomalous node. For example, a controller 122 can review workload results by a main node and redundant nodes, compare workload results, and if a majority of results are the same, then any different result is considered to be an anomaly. Please note main node and one redundant node correspond to Applicant’s first and second device, respectively. Please note the workloads submitted to the redundant nodes are copies of the workload submitted to the main node and therefore corresponds to Applicant’s copy of the partition. Please note a redundant node performing the workload corresponds to Applicant’s second device executing a copy of the partition.).
Li-Bajaj and Fargo are in the same field of endeavor in terms of computer systems and therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the teachings of Li-Bajaj to incorporate the teachings of Fargo wherein a first device in the plurality of devices executes a partition and a second device in the plurality of devices executes a copy of that partition. Having a plurality of devices processing the same data would decrease the chances of a delay or bottleneck from slower devices and speed up the entire system as a whole.
As per claim 12 Li-Bajaj teaches claim 11 as applied above but does not explicitly teach wherein a first device in the plurality of devices executes a partition of the partitioned neural network and a second device in the plurality of devices executes a copy of that partition.
However, Fargo teaches wherein a first device in the plurality of devices executes a partition and a second device in the plurality of devices executes a copy of that partition (e.g. Fargo: [0024] discloses after a workload is submitted by controller 122 to two or more of nodes 130-0 to 130-n, nodes will perform the workload and provide results that are accessible to controller 122. Controller 122 can collect the results and apply a voting mechanism technique to identify any anomalous node. For example, a controller 122 can review workload results by a main node and redundant nodes, compare workload results, and if a majority of results are the same, then any different result is considered to be an anomaly. Please note main node and one redundant node correspond to Applicant’s first and second device, respectively. Please note the workloads submitted to the redundant nodes are copies of the workload submitted to the main node and therefore corresponds to Applicant’s copy of the partition. Please note a redundant node performing the workload corresponds to Applicant’s second device executing a copy of the partition.).
Li-Bajaj and Fargo are in the same field of endeavor in terms of computer systems and therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the teachings of Li-Bajaj to incorporate the teachings of Fargo wherein a first device in the plurality of devices executes a partition and a second device in the plurality of devices executes a copy of that partition. Having a plurality of devices processing the same data would decrease the chances of a delay or bottleneck from slower devices and speed up the entire system as a whole.
Claims 3 and 13 are rejected under 35 U.S.C. 103 as being unpatentable over Li-Bajaj-Fargo in view of Prakash et al (US Pat. Pub. 2019/0220703; hereinafter referred to as Prakash).
As per claim 3 Li-Bajaj-Fargo teaches claim 1 and claim 2 as applied above but does not explicitly teach wherein the first device is located along the particular path and the second device is located along another path from among the different processing paths.
However, Prakash teaches wherein the first device is located along the particular path and the second device is located along another path from among the different processing paths (e.g. Prakash: [0050] discloses the MEC system 200 determines a coding redundancy to encode the training dataset X to mitigate wait time and bottleneck issues due to straggler nodes as discussed previously. The MEC system 200 determines the coding redundancy based on the operational parameters of each edge compute node 101, 201, which is used to encode the respective training datasets x.sub.1-x.sub.m. [Figure 1] discloses that the 101 and 201 nodes are different and follow separate paths.).
Li-Bajaj-Fargo and Prakash are in the same field of endeavor in terms of machine learning and therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the teachings of Li-Bajaj-Fargo to incorporate the teachings of Prakash wherein the first device is located along the particular path and the second device is located along another path from among the different processing paths. Having the devices located along different paths increases fault tolerance and improves speed by decreasing the chance of bottlenecks.
As per claim 13 Li-Bajaj-Fargo teaches claim 11 and claim 12 as applied above but does not explicitly teach wherein the first device is located along the particular path and the second device is located along another path from among the different processing paths.
However, Prakash teaches wherein the first device is located along the particular path and the second device is located along another path from among the different processing paths (e.g. Prakash: [0050] discloses the MEC system 200 determines a coding redundancy to encode the training dataset X to mitigate wait time and bottleneck issues due to straggler nodes as discussed previously. The MEC system 200 determines the coding redundancy based on the operational parameters of each edge compute node 101, 201, which is used to encode the respective training datasets x.sub.1-x.sub.m. [Figure 1] discloses that the 101 and 201 nodes are different and follow separate paths.).
Li-Bajaj-Fargo and Prakash are in the same field of endeavor in terms of machine learning and therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the teachings of Li-Bajaj-Fargo to incorporate the teachings of Prakash wherein the first device is located along the particular path and the second device is located along another path from among the different processing paths. Having the devices located along different paths increases fault tolerance and improves speed by decreasing the chance of bottlenecks.
Claims 6 and 16 are rejected under 35 U.S.C. 103 as being unpatentable over Li in view of Murahari et al (DataMUX: Data Multiplexing for Neural Networks; published February 2022; hereinafter referred to at Murahari).
As per claim 6 Li-Bajaj teaches claim 1 as applied above. Li-Bajaj does not teach wherein the plurality of devices executes at least one multiplexer and demultiplexer that connect two or more partitions of the partitioned neural network.
However, Murahari does teach wherein the plurality of devices executes at least one multiplexer and demultiplexer that connect two or more partitions of the partitioned neural network (e.g. Murahari: Page 3 [0004] discloses a multiplexer module to combine the multiple input instances into a superposed representation, a neural network backbone to process mixed representations, and a demultiplexing module to disentangle the processed representations for individual prediction.).
Li-Bajaj and Murahari are in the same field of endeavor in terms of neural networks and therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the teachings of Li-Bajaj to incorporate the teachings of Murahari to execute at least one multiplexer and demultiplexer that connect two or more partitions of the partitioned neural network. This would allow the processing of many instances simultaneously with minimal overhead during inference.
As per claim 16, Li-Bajaj teaches claim 11 as applied above. Li-Bajaj does not teach wherein the plurality of devices executes at least one multiplexer and demultiplexer that connect two or more partitions of the partitioned neural network.
However, Murahari does teach wherein the plurality of devices executes at least one multiplexer and demultiplexer that connect two or more partitions of the partitioned neural network (e.g. Murahari: Page 3 [0004] discloses a multiplexer module to combine the multiple input instances into a superposed representation, a neural network backbone to process mixed representations, and a demultiplexing module to disentangle the processed representations for individual prediction.).
Li-Bajaj and Murahari are in the same field of endeavor in terms of neural networks and therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the teachings of Li-Bajaj to incorporate the teachings of Murahari to execute at least one multiplexer and demultiplexer that connect two or more partitions of the partitioned neural network. This would allow the processing of many instances simultaneously with minimal overhead during inference.
Claim 10 is/are rejected under 35 U.S.C. 103 as being unpatentable over Li-Bajaj in view of Jia (US Pat. Pub. 2023/0129235).
As per claim 10, Li-Bajaj teaches claim 1 as applied above. Li-Bajaj does not teach the task request requests that the partitioned neural network make an inference about sensor data included in the task request.
However, Jia does teach the task request requests an inference about sensor data included in the task request (e.g. Jia: [0055] discloses in some embodiments, an inference can be formed by a machine learning component, an artificial intelligence component, etc., which inference can be employed in determining, designating, sorting, ordering, ranking, etc. [0088] discloses that the term “inference” can generally refer to the process of reasoning about, or inferring states of, the system, environment, user, and/or intent from a set of observations as captured via events and/or data. Captured data and events can include user data, device data, environment data, data from sensors, and sensor data.).
Li-Bajaj and Jia are in the same field of endeavor in terms of service differentiation and therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the teachings of Li-Bajaj to incorporate the teachings of Jia to have the task request requests an inference about sensor data included in the task request. By doing this, a second request will not have to be sent to specifically ask for an inference which will save on resources and time.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
US20220327359 (Ahuja, 10/13/2022): Levels of service (e.g. Ahuja: [0037] discloses to achieve results with low latency, the services executed within the Edge cloud 110 balance varying requirements in terms of: (a) Priority (throughput or latency) and Quality of Service (QoS) (e.g., traffic for an autonomous car may have higher priority than a temperature sensor in terms of response time requirement; or, a performance sensitivity/bottleneck may exist at a compute/accelerator, memory, storage, or network resource, depending on the application)).
CN202410009015 (Fan, priority date 01/03/2024): wherein a first device in the plurality of devices executes a partition of the partitioned neural network and a second device in the plurality of devices executes a copy of that partition (e.g. Fan: [0003] discloses : a main control daemon in a central control node determines a plurality of target computing nodes corresponding to tasks to be executed according to the tasks to be executed sent by a user, forms a computing node list according to the target computing nodes, and sends the tasks to be executed to daemons of the target computing nodes; the daemon process in the target computing node is used for running the task to be executed and sending the running result to the central control node; if the task process on the first target computing node in the computing node list is finished, the master control daemon transmits a clearing request to each target node so that the daemon in each target computing node clears the task to be executed.).
US20110302583 (Abadi, published 12/08/2011): wherein a first device in the plurality of devices executes a partition of the partitioned neural network and a second device in the plurality of devices executes a copy of that partition (e.g. Abadi: [0068] discloses the data processing task is portioned into a plurality of partitions. Each partition of the data processing task is assigned to a database system in the plurality of database systems for processing. [0128] further discloses Hadoop achieved fault tolerance by restarting tasks of failed nodes on other nodes. Please note failed nodes and other nodes correspond to Applicant’s first device and second device, respectively. Please note restarting tasks on a different node corresponds to Applicant’s second device executing a copy of the partition.).
Any inquiry concerning this communication or earlier communications from the examiner should be directed to DANIEL N TRUONG whose telephone number is (571)270-0856. The examiner can normally be reached Monday-Thursday 9:00AM-6:00PM.
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, April Blair can be reached at (571) 270-1014. 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.
/DANIEL NHU TRUONG/Examiner, Art Unit 2196
/APRIL Y BLAIR/Supervisory Patent Examiner, Art Unit 2196
Prosecution Insights