arXiv：cs.LG· Sen Lu, Xiaoyu Zhang, Mingtao Hu, Eric Yeu-Jer Lee, Soohyeon Kim, Wei D. Lu·· 7 小时前
State Space Models 自然产生时间细胞与振荡行为，并扩展至抽象认知功能
State Space Models Naturally Produce Time Cell and Oscillatory Behaviors and Scale to Abstract Cognitive Functions
AI 导读
一项基于 State Space Models 的研究表明，结构化的多时间尺度复数旋转基可在训练前自然生成类似时间细胞、斜坡活动与振荡响应的时序动态，并实现优化的历史压缩。该模型还能扩展至事件计数等抽象认知功能，为理解神经活动提供可计算的框架。
正文
Abstract:A grand challenge in modern neuroscience is bridging the gap between the microscale neural circuits and cognitive functions. While extensive knowledge exists about neuronal connectivity and biophysics, how low-level phenomena give rise to abstract behaviors remains largely unresolved. Here, we propose that a mathematical framework based on State Space Models offers mathematical sufficiency and computational plausibility to generate response profiles that resemble time cells, ramping activity, and oscillatory responses. Our results suggest that such behaviors can emerge from a structured multi-timescale complex rotational basis, which enables optimized history compression and naturally generates structured temporal dynamics even prior to training, mirroring recent findings in biological circuits. Learning exploits and adapts to the pre-existing temporal basis for specific tasks. We show that the model can readily scale to abstract cognitive functions such as event counting, supporting its use as a computationally tractable framework for understanding neural activities.
| Comments: | Sen Lu and Xiaoyu Zhang contributed equally. Wei D. Lu is the corresponding author. 5 figures are included in 15 pages |
| Subjects: | Neurons and Cognition (q-bio.NC); Machine Learning (cs.LG) |
| Cite as: | arXiv:2507.13638 [q-bio.NC] |
| (or arXiv:2507.13638v4 [q-bio.NC] for this version) | |
| https://doi.org/10.48550/arXiv.2507.13638 arXiv-issued DOI via DataCite |
Submission history
From: Sen Lu [view email]
[v1]
Fri, 18 Jul 2025 03:53:16 UTC (1,088 KB)
[v2]
Thu, 22 Jan 2026 19:09:45 UTC (1,236 KB)
[v3]
Tue, 10 Feb 2026 05:35:53 UTC (1,515 KB)
[v4]
Wed, 14 Oct 2026 12:48:42 UTC (2,205 KB)
来源：arXiv：cs.LG · arxiv.org