Storage medium, optimum solution acquisition method and information processing apparatus

Patent metadata

Field	Value
Publication number	US-12051003-B2
Application number	US-202017027757-A
Country	US
Kind code	B2
Filing date	Sep 22, 2020
Priority date	Oct 1, 2019
Publication date	Jul 30, 2024
Grant date	Jul 30, 2024

Abstract

A non-transitory computer-readable storage medium storing a program that causes a computer to execute a process, the process includes obtaining a machine learning model having learned characteristic amounts of a plurality of training data including an objective function; calculating similarities between the characteristic amounts of the plurality of training data by inputting the plurality of training data to the obtained machine learning model; specifying a data group having a high similarity with a desired objective function from the characteristic amounts of the plurality of training data based on distances of the calculated similarities; and acquiring an optimum solution for the desired objective function by using the specified data group.

First claim

What is claimed is:

1. A non-transitory computer-readable storage medium storing a program that causes a computer to execute a process, the process comprising:
training a variational autoencoder using a plurality of training data including objective functions and parameters having an influence on the objective functions;
generating a solution space in which parts with first objective functions and parts with second objective functions less than the first objective functions are placed, respectively, in a concentrated manner over a latent space of a variational autoencoder by inputting the plurality of training data to the trained variational autoencoder;
calculating a set of objective functions, variables and characteristic values by inputting a sampling set generated from the latent space to a decoder of the variational autoencoder;
acquiring a lowest value or a highest value of the objective functions of the set as an optimum solution for a desired objective function;
outputting, when acquiring the optimum solution, the set of objective functions, variables and characteristic values; and
generating, when not acquiring the optimum solution, training data for re-learning by changing a fluctuation range of the variables of the set.

2. The non-transitory computer-readable storage medium storing a program according to claim 1 , wherein the acquiring includes acquiring a parameter giving the optimum solution for the desired objective function by using the decoder of the variational autoencoder.

3. An optimum solution acquisition method executed by a computer, the optimum solution acquisition method comprising:
training a variational autoencoder using a plurality of training data including objective functions and parameters having an influence on the objective functions;
generating a solution space in which parts with first objective functions and parts with second objective functions less than the first objective functions are placed, respectively, in a concentrated manner over a latent space of a variational autoencoder by inputting the plurality of training data to the trained variational autoencoder;
calculating a set of objective functions, variables and characteristic values by inputting a sampling set generated from the latent space to a decoder of the variational autoencoder;
acquiring a lowest value or a highest value of the objective functions of the set as an optimum solution for a desired objective function;
outputting, when acquiring the optimum solution, the set of objective functions, variables and characteristic values; and
generating, when not acquiring the optimum solution, training data for re-learning by changing a fluctuation range of the variables of the set.

4. The optimum solution acquisition method according to claim 3 , wherein the acquiring includes acquiring a parameter giving the optimum solution for the desired objective function by using the decoder of the variational autoencoder.

5. An information processing apparatus, comprising:
a memory; and
a processor coupled to the memory and configured to:
train a variational autoencoder using a plurality of training data including objective functions and parameters having an influence on the objective functions,

generate a solution space in which parts with first objective functions and parts with second objective functions less than the first objective functions are placed, respectively, in a concentrated manner over a latent space of a variational autoencoder by inputting the plurality of training data to the trained variational autoencoder,
calculate a set of objective functions, variables and characteristic values by inputting a sampling set generated from the latent space to a decoder of the variational autoencoder,
acquire a lowest value or a highest value of the objective functions of the set as an optimum solution for a desired objective function,
output, when acquiring the optimum solution, the set of objective functions, variables and characteristic values, and
generate, when not acquiring the optimum solution, training data for re-learning by changing a fluctuation range of the variables of the set.

Assignees

• Fujitsu Ltd

Inventors

• Ohta Eiji
• Akima Hisanao

Classifications

• G06N3/0455

  Auto-encoder networks; Encoder-decoder networks · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06N3/0475

  Generative networks · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N3/045

  Combinations of networks · CPC title