• Search Research Projects
  • Search Researchers
  • How to Use
  1. Back to project page

2022 Fiscal Year Final Research Report

Interactive visualization for verification of training data for machine learning

Research Project

  • PDF
Project/Area Number 20K11917
Research Category

Grant-in-Aid for Scientific Research (C)

Allocation TypeMulti-year Fund
Section一般
Review Section Basic Section 61020:Human interface and interaction-related
Research InstitutionOchanomizu University

Principal Investigator

Takayuki Itoh  お茶の水女子大学, 基幹研究院, 教授 (80401595)

Project Period (FY) 2020-04-01 – 2023-03-31
Keywords可視化 / 訓練データ / アノテーション
Outline of Final Research Achievements

This research focused on the depelopment of various methods to support the construction of high-quality training data for machine learning, by visualizing the distribution and creation process of the training data. As the results, we proposed the following three types of visualization methods: 1) Semi-automation of the annotation process of training data and visualization of the results of the construction and operation of the decision tree that serves as the basis for the annotation process; 2) Comparative visualization of distributions of features and labels of multiple training data; and 3) Visualization of the annotation process by multiple workers and verification of the reliability of the annotations.

Free Research Field

可視化

Academic Significance and Societal Importance of the Research Achievements

大規模で複合的な訓練データの分布や制作過程を視認性の高い形で情報提示する手法の開発は、可視化の研究における学術面での本質的な課題であり、これを解くことに学術的意義があった。一方で、機械学習の普及により訓練データの品質は社会的に大きな課題となっている。訓練データ制作の半自動化による信頼性の向上、複数の訓練データ間での特徴量やラベルの分布の検証、訓練データの制作過程での各作業者による工程の信頼性の検証、といった各課題は機械学習の品質を向上するために重要な課題であり、これらの解決には大きな社会的意義があった。

URL: 

Published: 2024-01-30  

Information User Guide FAQ News Terms of Use Attribution of KAKENHI

Powered by NII kakenhi