Data dimensionality reduction method, computer program, and data dimensionality reduction device

Patent metadata

Field	Value
Publication number	US-12253993-B2
Application number	US-202218573981-A
Country	US
Kind code	B2
Filing date	Jun 10, 2022
Priority date	Jun 23, 2021
Publication date	Mar 18, 2025
Grant date	Mar 18, 2025

Abstract

A data dimensionality reduction method includes: a step of dimensionally reducing a group of data from a high-dimensional space to a low-dimensional space using a distance function that defines a distance between any two vectors in the high-dimensional space; a step of dividing the dimensionally-reduced low-dimensional space into multiple subspaces; an analysis step of performing a regression analysis using a regression model based on at least one belonging data for each divided subspace; and a step of updating p first parameters included in the distance function based on results of the regression analysis in the multiple subspaces.

First claim

The invention claimed is:

1. A method for dimensionally reducing a group of data each represented by an m-dimensional vector in an m-dimensional (m is an integer of 3 or more) high-dimensional space to an n-dimensional (n is an integer of 2 or more and less than m) low-dimensional space, the method comprising:
a reduction step of dimensionally reducing the group of data from the high-dimensional space to the low-dimensional space using a distance function that defines a distance between any two vectors in the high-dimensional space, wherein the distance function includes p (p is an integer of m or more) first parameters;
a division step of dividing the dimensionally-reduced low-dimensional space into multiple subspaces;
an analysis step of performing a regression analysis using a regression model based on at least one belonging data for each divided subspace, wherein the regression model is represented as a function of m explanatory variables and q (q is an integer of m or more) second parameters corresponding to the m explanatory variables; and
an update step of updating the p first parameters included in the distance function based on results of the regression analysis in the multiple subspaces,
wherein the reduction step, the division step, the analysis step and the update step are repeatedly performed.

2. The method according to claim 1 , wherein in the reduction step, a self-organizing map is applied to the dimension reduction.

3. The method according to claim 1 , wherein:
the distance function is represented by a weighted Euclidean distance, where the integer p is equal to the integer m; and
the p first parameters are m weight coefficients that weigh a distance between m components between the any two vectors.

4. The method according to claim 1 , wherein the regression model is a linear multiple regression model.

5. The method according to claim 4 , wherein:
the q second parameters are m partial regression coefficients, where the integer q is equal to the integer m;
an update parameter is calculated based on absolute values of the m partial regression coefficients obtained from a linear multiple regression analysis performed in the analysis step for each divided subspace; and
in the update step, the p first parameters are updated using the update parameter.

6. The method according to claim 1 , wherein the low-dimensional space is a 2-dimensional space.

7. The method according to claim 1 , comprising an initialization step of initializing the p first parameters.

8. The method according to claim 1 , comprising a display step of displaying on a display device an image visualizing a group of data in the dimensionally-reduced low-dimensional space.

9. The method according to claim 1 , wherein the m-dimensional vector includes, as its components, a chemical composition and/or manufacturing conditions of an alloy material.

10. The method according to claim 9 , wherein the manufacturing conditions include manufacturing conditions associated with a rolled material of an aluminum alloy.

11. A computer program, stored on a non-transitory computer readable storage medium, used for dimensionally reducing a group of data each represented by an m-dimensional vector in an m-dimensional (m is an integer of 3 or more) high-dimensional space to an n-dimensional (n is an integer of 2 or more and less than m) low-dimensional space, the computer program causing a computer to perform:
a reduction step of dimensionally reducing the group of data from the high-dimensional space to the low-dimensional space using a distance function that defines a distance between any two vectors in the high-dimensional space, wherein the distance function includes p (p is an integer of m or more) first parameters;
a division step of dividing the dimensionally-reduced low-dimensional space into multiple subspaces;
an analysis step of performing a regression analysis using a regression model based on at least one belonging data for each divided subspace, wherein the regression model is represented as a function of m explanatory variables and q (q is an integer of m or more) second parameters corresponding to the m explanatory variables; and
an update step of updating the p first parameters included in the distance function based on results of the regression analysis in the multiple subspaces,
wherein the computer program causes the computer to repeatedly perform the reduction step, the division step, the analysis step and the update step in this order.

12. A data dimensionality reduction device for dimensionally reducing a group of data each represented by an m-dimensional vector in an m-dimensional (m is an integer of 3 or more) high-dimensional space to an n-dimensional (n is an integer of 2 or more and less than m) low-dimensional space, the data dimensionality reduction device comprising:
a processor; and
a memory storing a program for controlling an operation of the processor, the processor performing, according to the program:
a reduction step of dimensionally reducing the group of data from the high-dimensional space to the low-dimensional space using a distance function that defines a distance between any two vectors in the high-dimensional space, wherein the distance function includes p (p is an integer of m or more) first parameters;
a division step of dividing the dimensionally-reduced low-dimensional space into multiple subspaces;
an analysis step of performing a regression analysis using a regression model based on at least one belonging data for each divided subspace, wherein the regression model is represented as a function of m explanatory variables and q (q is an integer of m or more) second parameters corresponding to the m explanatory variables; and
an update step of updating the p first parameters included in the distance function based on results of the regression analysis in the multiple subspaces,
wherein the processor repeatedly performs the reduction step, the division step, the analysis step and the update step in this order.

Assignees

• Uacj Corp

Inventors

• Maeno Ryota

Classifications

• G06F16/23

  Updating · CPC title

• G06F16/2237

  Vectors, bitmaps or matrices · CPC title

• Y02P90/30

  Computing systems specially adapted for manufacturing · CPC title

• G06N20/00

  Machine learning · CPC title

• G06F16/2264 Primary

  Multidimensional index structures · CPC title