arxiv.org/pdf/2106.05241… "We point out that this run is cherry-picked from over 100 runs with different random initialisation." I really dig they discussed the issue of stability in Section E. It's been a lonely battle against algo. stability in unsupervised disentangling models.

Hi ISMIR Conference people! I'm going to present this in the afternoon. Feel free to stop by and come to our poster or ask questions in the slack channel p6-01-yu. See you there!

Thank you for visiting our poster in the LBD session 🎶 #ISMIR2023

2/2 first authored papers accepted to ICASSP! Both on improving disentangled sequential autoencoders applied to musical instrument sounds and speech. Completely unsupervised. Looking forward to 🇰🇷🍖🍶 and of course meeting VAE folks!

Sad how we audio people are unexposed to the CV folks. We have attempted the challenges listed in this ICML24 openreview.net/pdf?id=AocOA4h… - difficulty in tuning latent dimension - requirement of domain-dependent data augmentation - overheads by auxiliary losses such as mutual info

Audio codecs are turning into VQ-based voice conversion models with an extra focus on compression. Disentanglement seems to be the sauce.

``Self-Supervised Multi-View Learning for Disentangled Music Audio Representations,'' Julia Wilkins, Sivan Ding, Magdalena Fuentes, Juan Pablo Bello, ift.tt/NFIb2yd

There goes the one proof to my ISMIR presence. It was very nice to catch up w/ the Taiwanese Gang, and it's my honour to be confronted by "why are you still doing disentanglement?" That's right, I will also be presenting DisMix openreview.net/pdf?id=Zhc6ZFm… at the NeurIPS Workshop!

主著論文がAPSIPA Trans.にアクセプトされました🙌 Our paper has been accepted for publication in APSIPA Transactions!!🚀 A big thanks to the co-authors (professors!), reviewers, and everyone who supported this work. Special mention to Yin-Jyun Luo, たいし, and Yoshiaki Bando🙏

Pumped to see a comeback of GMVAE among a sea of VQ! openreview.net/forum?id=cuFzE… Speaking of, Wei-Ning's arxiv.org/abs/1810.07217 on TTS has a substantial impact to my research on style transfer via (unsupervised) disentanglement. But it seems overshadowed by his own work HuBERT😅

Preprint of today: Vavilala et al., "Generative Blocks World: Moving Things Around in Pictures" -- arxiv.org/abs/2506.20703 I have a soft spot for reviving old ideas in modern methods -- block world via primitives now with Diffusion models for generating/editing images.