Information processing apparatus, information processing method, information processing program, and information processing system

Patent metadata

Field	Value
Publication number	US-2025202786-A1
Application number	US-202318843318-A
Country	US
Kind code	A1
Filing date	Mar 13, 2023
Priority date	Mar 14, 2022
Publication date	Jun 19, 2025
Grant date	—

Abstract

An information processing apparatus according to an embodiment includes: a controller that selects, from a plurality of networks, one or more first networks executed on a one-to-one basis by one or more first processors different from each other and selects, from the plurality of networks, a second network executed by a second processor; and a transmission unit that transmits each of the one or more first networks to each of the one or more first processors on a one-to-one basis and transmits the second network to the second processor. The second processor executes the second network by using, as an input, output data that is output as a result of executing a network selected from the one or more first networks for at least one processor among the one or more first processors. The controller selects the second network from the plurality of networks according to the output data.

First claim

1 . An information processing apparatus comprising:
a controller configured to select, from a plurality of networks, one or more first networks executed on a one-to-one basis by one or more first processors different from each other, and select, from the plurality of networks, a second network executed by a second processor; and a transmission unit configured to transmit the one or more first networks to the one or more first processors on a one-to-one basis, and transmit the second network to the second processor, wherein the second processor executes the second network using output data as an input, the output data being output as a result of executing a network selected from the one or more first networks for at least one processor among the one or more first processors, and the controller selects, from the plurality of networks, the second network according to the output data.

2 . The information processing apparatus according to claim 1 , wherein
at least one of the one or more first processors is different from the second processor.

3 . The information processing apparatus according to claim 1 , wherein
the controller adaptively selects the second network according to the output data.

4 . The information processing apparatus according to claim 1 , wherein
the controller determines the one or more first networks and the second network according to required performance that is performance required for execution of processing by at least one of the one or more first networks and the second network.

5 . The information processing apparatus according to claim 4 , wherein
the controller determines the one or more first networks and the second network based on at least one of the required performance for a hardware element and the required performance for an application element, the application element being required for the output data by an application that executes processing on the output data.

6 . The information processing apparatus according to claim 5 , wherein
the hardware element includes capability and constraint of at least one of the one or more first processors, the second processor, and hardware generating input data to be input to the one or more first networks, and the application element includes at least one of a rate of the output data, a transmission size, and a processing cost in the application.

7 . The information processing apparatus according to claim 4 , wherein
the controller determines the one or more first networks and the second network based on the required performance and a key performance indicator for a predetermined network.

8 . The information processing apparatus according to claim 4 , wherein
the controller determines the one or more first networks and the second network so as to perform a fixed point operation in the one or more first networks and a floating point operation in the second network.

9 . The information processing apparatus according to claim 4 , wherein
the controller determines processing by the one or more first networks and processing by the second network such that the processing by the second network is executed as a pipeline process on the processing by each of the one or more first networks.

10 . The information processing apparatus according to claim 1 , wherein
one or more first functions for input data is applied to the one or more first networks on a one-to-one basis, the input data being input to the one or more first networks included in a predetermined network, and a second function for a processing result of the first function is applied to the second network.

11 . The information processing apparatus according to claim 10 , wherein
the controller determines each of the one or more first functions according to the second function.

12 . The information processing apparatus according to claim 10 , wherein
the controller updates or changes the second function according to a learning result of the second network.

13 . The information processing apparatus according to claim 10 , wherein
the controller determines, as the second network, a plurality of the second networks that corresponds on a one-to-one basis to a plurality of the second functions different from each other and is capable of parallel processing.

14 . The information processing apparatus according to claim 13 , wherein
the controller determines the plurality of second networks so as to commonly input an output of one of the one or more first networks.

15 . The information processing apparatus according to claim 10 , wherein
as the one or more first networks, a plurality of the first networks that corresponds on a one-to-one basis to the one or more first functions different from each other and is capable of parallel processing is determined.

16 . The information processing apparatus according to claim 15 , wherein
the controller determines the second network so as to receive an output of each of the one or more first networks.

17 . The information processing apparatus according to claim 10 , wherein
the controller determines each of the one or more first networks giving priority to speed with respect to processing of each of the one or more first functions, and determines the second network giving priority to at least one of accuracy and a function with respect to processing of the second function.

18 . The information processing apparatus according to claim 17 , wherein
the controller determines the one or more first networks and the second network so as to retrain the one or more first networks based on a processing result of the second network.

19 . The information processing apparatus according to claim 10 , wherein
the controller determines the one or more first networks and the second network so as to reduce data output from the one or more first networks and input the data reduced to the second network.

20 . An information processing method executed by a computer, the information processing method comprising:
a control step of selecting, from a plurality of networks, one or more first networks executed on a one-to-one basis by one or more first processors different from each other, and selecting, from the plurality of networks, a second network executed by a second processor; and a transmitting step of transmitting the one or more first networks to the one or more first processors on a one-to-one basis and transmitting the second network to the second processor, wherein the second processor executes the second network using output data as an input, the output data being output as a result of executing a network selected from the one or more first networks for at least one processor among the one or more first processors, and the control step includes selecting, from the plurality of networks, the second network according to the output data.

21 . An information processing program causing a computer to execute:
a control step of selecting, from a plurality of networks, one or more first networks executed on a one-to-one basis by one or more first processors different from each other, and selecting, from the plurality of networks, a second network executed by a second processor; and a transmitting step of transmitting the one or more first networks to the one or more first processors on a one-to-one basis and transmitting the second network to the second processor, wherein the second processor executes the second network using output data as an input, the

Assignees

• Sony Semiconductor Solutions Corp

Inventors

• Yamada Kazuki
• Yang Ying
• Jagadeesh Hareesh Gowtham
• Ishida Minoru
• Okumura Hiroyuki
• Okuike Kazuyuki

Classifications

• H04L41/16

  using machine learning or artificial intelligence · CPC title

• H04L41/082

  the condition being updates or upgrades of network functionality · CPC title

• G06T7/00

  Image analysis · CPC title

• G06N3/098

  Distributed learning, e.g. federated learning · CPC title

• G06F9/5055

  considering software capabilities, i.e. software resources associated or available to the machine · CPC title