Recipient of the Best Paper Award at the 2026 Summer Comprehensive Academic Conference of the Korean Society for Information Technology / Baek Seung-hoon (Master’s student, Department of Data Science, Graduate School, Class of ’24)
- 26.08.06 / 홍유민
Baek Seung-hoon, a master’s student in the Department of Data Science at Kookmin University’s Graduate School (advisor: Professor Lee Jae Hyuk), received the Outstanding Paper Award at the “2026 Summer Comprehensive Academic Conference of the Korean Society for Information Technology,” held from June 4 to 6, 2026.
Baek Seung-hoon presented a paper titled “Learning Differentiable Tokenizers for Self-Supervised Audio Anomaly Detection” at the conference. This research focused on the discrete token-based approach, one of the methods used in audio self-supervised learning. In particular, to address the gradient discontinuity issue that can occur during the generation of discrete tokens in the BRIDLE (Bidirectional Residual Quantization Interleaved Discrete Learning Encoder) model, the study proposed a differentiable tokenizer learning method utilizing Gumbel-Softmax. The research team applied the proposed method to various audio datasets to compare and analyze performance, thereby confirming the potential for improved training efficiency and performance in audio anomaly detection models.
This award is significant in that it demonstrates the potential of new learning methods in the fields of audio self-supervised learning and anomaly detection.
|
This content is translated from Korean to English using the AI translation service DeepL and may contain translation errors such as jargon/pronouns. If you find any, please send your feedback to kookminpr@kookmin.ac.kr so we can correct them.
|
|
Recipient of the Best Paper Award at the 2026 Summer Comprehensive Academic Conference of the Korean Society for Information Technology / Baek Seung-hoon (Master’s student, Department of Data Science, Graduate School, Class of ’24) |
||||
|---|---|---|---|---|
|
Baek Seung-hoon, a master’s student in the Department of Data Science at Kookmin University’s Graduate School (advisor: Professor Lee Jae Hyuk), received the Outstanding Paper Award at the “2026 Summer Comprehensive Academic Conference of the Korean Society for Information Technology,” held from June 4 to 6, 2026. Baek Seung-hoon presented a paper titled “Learning Differentiable Tokenizers for Self-Supervised Audio Anomaly Detection” at the conference. This research focused on the discrete token-based approach, one of the methods used in audio self-supervised learning. In particular, to address the gradient discontinuity issue that can occur during the generation of discrete tokens in the BRIDLE (Bidirectional Residual Quantization Interleaved Discrete Learning Encoder) model, the study proposed a differentiable tokenizer learning method utilizing Gumbel-Softmax. The research team applied the proposed method to various audio datasets to compare and analyze performance, thereby confirming the potential for improved training efficiency and performance in audio anomaly detection models. This award is significant in that it demonstrates the potential of new learning methods in the fields of audio self-supervised learning and anomaly detection.
|
||||
