KEN NISHIDA

Research Overview: Multi-label Classification

Multi-label classification is a task that classifies data points that may be assigned multiple labels simultaneously. Unlike simple classification problems with a single label, multi-label classification can have multiple labels associated with each data point.

For example, in image classification, multi-label classification assigns multiple labels to an image when it represents different objects or attributes simultaneously. For instance, an image of a dog might be labeled with both "dog" and "walking."

Multi-label classification is used in various applications. For instance, in text classification, it is used when a document is related to multiple topics. It is also applied in music and video classification to predict labels associated with multiple genres or attributes.

Publications

• Ken Nishida, Kojiro Machi, Kazuma Onishi, Katsuhiko Hayashi, Hidetaka Kamigaito. Multi-label Learning with Random Circular Vectors. arXiv, July 2024 [Paper]
• Ken Nishida, Katsuhiko Hayashi, Kojiro Machi, Hidetaka Kamigaito. Multi-label Classification using Circular Vectors. IPSJ 258th NL Research Meeting, December 2023 [Paper] [Slide]
• Ken Nishida, Katsuhiko Hayashi, Hidetaka Kamigaito. Efficient Large-scale Multi-label Classification with Circular Holographic Reduced Representations. 18th NLP Young Researchers Symposium (YANS), August 2023 [Poster]