Fundamental algorithms for detecting music similarities from various viewpoints

2016 Fiscal Year Final Research Report

• Project/Area Number: 26330243
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Intelligent informatics
• Research Institution: Hokkaido University
• Principal Investigator: OKUBO Yoshiaki 北海道大学, 情報科学研究科, 助教 (40271639)
• Project Period (FY): 2014-04-01 – 2017-03-31
• Keywords: 楽曲類似検索 / クラスタ / 形式概念分析 / 疑似クリーク / 分類問題

Outline of Final Research Achievements

The purpose of this research is to investigate methods for flexible and useful Music Information Retrieval. Especially, we have tried to formalize several fundamental methods for observing and analyzing similar music from various viewpoints. The algorithms we have developed based on Pattern Mining (Formal Concept Analysis) and Dense-Subgraph Discovery are 1) Algorithm for similar music retrieval based on colossal frequent patterns, and 2) Algorithm for extracting Top-N largest dense subgraphs based on proper k-plex search. Furthermore, we also have developed 3) Classification algorithm with label modification based on multiple majority votes, and 4) Classification algorithm with dimension reduction based on class-label information.

Free Research Field

知能情報学