講演情報
[1Yin-A-49]Tensor Brain: Structured Probabilistic Modeling for Neural Population Activity with Tensor Networks
〇Yunyu Huang1, Fujia Wu1, Dairui Chen1, Xiaowei Gu4, Zhe Sun2,3, Chao LI2 (1. Graduate School of Health Data Science, Juntendo University, 2. Faculty of Health Data Science, Juntendo University, 3. Graduate School of Medicine, Juntendo University, 4. RIKEN Center for Brain Science)
キーワード：
Probabilistic modeling、Generative models、Tensor networks、Neural population activity、Time-series modeling
Probabilistic modeling of neural population activity requires representing the high-dimensional joint distribution of coordinated spike trains across neurons and time. However, the combinatorial growth of the joint state space makes direct modeling computationally intractable. We propose Tensor Brain (TB), a tensor-network-based probabilistic framework that factorizes the full joint distribution into structured tensor contractions. TB enables scalable joint modeling without introducing latent variables and supports exact likelihood evaluation. To ensure numerical stability in long tensor contractions, we derive a scaled contraction scheme and optimize core tensors under unitary constraints on the Stiefel manifold. Experiments on synthetic data, calcium imaging recordings, and the MC_Maze dataset show that TB accurately captures spike-count statistics, pairwise correlations, and temporal dependencies. These results demonstrate that TB provides a scalable and principled framework for modeling large-scale neural population activity.