A Disentangled Low-Rank RNN Framework for Uncovering Neural Connectivity and Dynamics

2025-11-17

AI Summary

Low-rank recurrent neural networks (lrRNNs) effectively capture low-dimensional dynamics of neural population activity but suffer from entangled functional roles of latent dimensions and lack interpretable, functionally decoupled connectivity. To address this, we propose DisRNN—a novel framework that enforces inter-group independence and intra-group flexible coupling within a low-rank recurrent architecture, augmented with partial correlation regularization to achieve functional disentanglement of latent variables. Implemented within a variational autoencoder (VAE) framework, DisRNN is jointly trained and validated on both synthetic data and real neural recordings (macaque M1 spiking data and mouse voltage imaging). Experiments demonstrate that DisRNN significantly improves both the functional decoupling of latent trajectories and the neuroscientific interpretability of low-rank connectivity, outperforming existing lrRNN methods. This work establishes a new paradigm for dissecting functional specialization in neural computation.

Technology Category

• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Machine Learning: Learning with Manifolds
• Reasoning under Uncertainty: Relational Probabilistic Models

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization

Abstract

Low-rank recurrent neural networks (lrRNNs) are a class of models that uncover low-dimensional latent dynamics underlying neural population activity. Although their functional connectivity is low-rank, it lacks disentanglement interpretations, making it difficult to assign distinct computational roles to different latent dimensions. To address this, we propose the Disentangled Recurrent Neural Network (DisRNN), a generative lrRNN framework that assumes group-wise independence among latent dynamics while allowing flexible within-group entanglement. These independent latent groups allow latent dynamics to evolve separately, but are internally rich for complex computation. We reformulate the lrRNN under a variational autoencoder (VAE) framework, enabling us to introduce a partial correlation penalty that encourages disentanglement between groups of latent dimensions. Experiments on synthetic, monkey M1, and mouse voltage imaging data show that DisRNN consistently improves the disentanglement and interpretability of learned neural latent trajectories in low-dimensional space and low-rank connectivity over baseline lrRNNs that do not encourage partial disentanglement.

Problem

• Low-rank RNNs lack disentanglement interpretations for neural connectivity
• Existing models cannot assign distinct computational roles to latent dimensions
• Current approaches struggle with interpretable neural dynamics in low-dimensional space

Innovation

• Disentangled RNN framework with group-wise independence
• Variational autoencoder with partial correlation penalty
• Separate latent group evolution for complex computation