DiVeQ: Differentiable Vector Quantization Using the Reparameterization Trick
Mohammad Hassan Vali
•
•
1 min readProfessional Summary
Mohammad Hassan Vali is a Postdoctoral Researcher in the Department of Computer Science at Aalto University, Finland. His postdoctoral research, supervised by Prof. Arno Solin and funded by the Finnish Center for Artificial Intelligence (FCAI), focuses on discrete representation learning in deep neural networks. In particular, his work explores efficient vector quantization techniques with applications in image and speech processing, including image compression and generation, 3D Gaussian splatting compression, interpretable AI, speech privacy, and voice conversion.
He earned his Ph.D. in Electrical Engineering from Aalto University in 2025, where his doctoral thesis, “Vector Quantization in Deep Neural Networks for Speech and Image Processing,” was supervised by Prof. Tom Bäckström. He received his B.S. and M.S. degrees in Electrical Engineering from Babol Noshirvani University, Iran, in 2014 and 2017, respectively.
Education
Ph.D.
Aalto University, Finland
M.Sc.
Babol Noshirvani University, Iran
B.Sc.
Babol Noshirvani University, Iran
Interests
Signal Processing
Machine Learning
Computer Vision
Vector Quantization
Privacy
Research Highlights
Read more
Unsupervised Panoptic Interpretation of Latent Spaces in GANs Using Space-Filling Vector Quantization
Mohammad Hassan Vali
•
•
1 min read
Interpretable Latent Space Using Space-Filling Curves for Phonetic Analysis in Voice Conversion
Mohammad Hassan Vali
•
•
1 min read
NSVQ: Noise Substitution in Vector Quantization for Machine Learning
Mohammad Hassan Vali
•
•
1 min readRecent Publications
(2026).
Smol-GS: Compact Representations for Abstract 3D Gaussian Splatting.
Submitted to Proceedings of the Neural Information Processing Systems (NeurIPS).
(2026).
DiVeQ: Differentiable Vector Quantization Using the Reparameterization Trick.
Proceedings of the International Conference on Learning Representations (ICLR).
(2026).
Self-Attention Decomposition for Training Free Diffusion Editing.
Proceedings of the IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP).
(2026).
Sparsely Supervised Diffusion.
Submitted to Proceedings of the Neural Information Processing Systems (NeurIPS).
(2025).
Privacy Disclosure of Similarity Rank in Speech and Language Processing.
IEEE Transactions on Audio, Speech and Language Processing.