Machine learning model construction device, numerical control, machine learning model construction method, and non-transitory computer readable medium encoded with a machine learning model construction program

Patent metadata

Field	Value
Publication number	US-2018024510-A1
Application number	US-201715648753-A
Country	US
Kind code	A1
Filing date	Jul 13, 2017
Priority date	Jul 22, 2016
Publication date	Jan 25, 2018
Grant date	—

Abstract

Various neural network models are constructed flexibly on a. learning circuit. A machine learning model construction device includes a learning circuit ( 80 ) capable of constructed a neural network model according to a setting value, and a control means ( 11 ) capable of adjusting the setting value so as to become a value for constructing a predetermined neural network model in the learning circuit ( 80 ).

First claim

1 . A machine learning model construction device comprising:
a learning circuit capable of constructing a neural network model according to a setting value; and a control means for adjusting the setting value so as to be a value for constructing a predetermined neural network model in the learning circuit.

2 . A machine learning model construction device according to claim 1 , wherein the control means receives input of a parameter value, defines a neural network model corresponding to the received parameter value as the predetermined neural network model, and adjusts the setting value so as to be a value for constructing the predetermined neural network model in the learning circuit.

3 . A machine learning model construction device according to claim 2 , wherein the received parameter value is a value designating a layer number of the neural network model and a dimension number of each layer.

4 . A machine learning model construction device according to claim 2 , wherein the control means reads the parameter value from a setting file.

5 . A machine learning model construction device according to claim 1 , wherein the control means draws a connection relationship of perceptrons according to the parameter value, and a graphical user interface for performing adjustment of the parameter value while a user references the connection relationship of the perceptrons, on a display means.

6 . A machine learning model construction device according to claim 1 , wherein the control means allows a user to select a neural network model prepared in advance, defines the neural network model selected as the predetermined neural network model, and adjusts the setting value so as to be a value for constructing the predetermined neural network model in the learning circuit.

7 . A machine learning model construction device according to claim 1 , wherein the control means constructs a neural network model in the learning circuit by way of controlling output of each perceptron included in the learning circuit according to the setting value.

8 . A machine learning model construction device according to claim 7 , wherein controlling output of each perceptron included in the learning circuit according to the setting value includes setting output of a multiplier within each perceptron included in the learning circuit according to the setting value, as any value among a value arrived at by multiplying a weighting value by input of the multiplier, a value as inputted to the multiplier, and a value of zero, and sets a sum of outputs of multipliers within each perceptron as an output of each perceptron.

9 . A numerical control for controlling a machine tool, comprising the machine learning model construction device according to claim 1 ,
wherein the control means receives input of a parameter value for constructing a neural network model that is related to control of the machine tool in the learning circuit, sets a neural network model corresponding to the received parameter value as the predetermined neural network model, and adjusts the setting value so as to be a value for constructing the predetermined neural network model in the learning circuit.

10 . A non-transitory computer readable medium encoded with a machine learning model construction program for causing a computer to function as a machine learning construction device, comprising:
a learning circuit capable of constructing a neural network model according to a setting value; and a control means for adjusting the setting value so as to be a value for constructing a predetermined neural network model in the learning circuit.

11 . A method of constructing a machine learning model performed by a machine learning model construction device including a learning circuit capable of constructing a neural network model according to a setting value, the method comprising:
a step of controlling to adjust the setting value so as to be a value for constructing a predetermined neural network model in the learning circuit.

Assignees

• Fanuc Corp

Inventors

• Matsushima Takahiko
• Miyazaki Yoshito

Classifications

• G06N3/063

  using electronic means · CPC title

• G05B13/027 Primary

  using neural networks only · CPC title