NeuroLens: Learning Latent Embeddings of Neural Semantics from Chronic Recordings
Abstract
Understanding how neural activity represents higher-order cognition and how these representations evolve over time has long been a central pursuit in neuroscience. However, current analytical tools cannot easily distinguish representational plasticity from recording instability in chronic neural recordings. Here, we introduce NeuroLens (Latent Embeddings of Neural Semantics), a self-supervised model based on the Joint-Embedding Predictive Architecture (JEPA) framework that learns denoised, semantically informative latents from chronic neural recordings. An adaptive encoder maps changing neural populations into a common latent space, while a temporal predictor learns structure that supports prediction of future latent states. By predicting in latent space, NeuroLens captures temporally predictable structure and reduces sensitivity to transient, recording-specific variability. Across chronic intracortical data in mice and humans, the learned representations improve decoding of decision-making and semantic task variables. Multi-day pretraining enables generalization to future sessions, rapid few-shot adaptation to unseen neural populations, and more stable decoding over time than state-of-the-art baselines. Together, these results establish NeuroLens as a new paradigm for studying how neural representations change during learning and over long timescales.
Related
- Protocol for chronic electroencephalography, fiber photometry, and silicon probe recordings in the Syrian hamster.
- GeoEPred: A multimodal structure-aware geometric deep learning framework for Gram-negative bacterial secreted effector prediction with sequence semantics
- Chemical Descriptors and Deep Learning Embeddings for Scoring de novo Peptide Designs
- MOSAIC: Learning Graph Node Embeddings from Spectrally Isolated Dominant Subspaces of Accumulated Diffusion Operators
- Interrogating contrastive learning embeddings for structure-based virtual screening: a case study on DrugCLIP
BibTeX
@misc{lyu2026neurolens,
title = {{NeuroLens: Learning Latent Embeddings of Neural Semantics from Chronic Recordings}},
author = {Hanrui Lyu and Baiyuan Chen and Tianshu Tan and Matthew R. Whiteway and Maxwell D. Melin and Ji Xia and Linyang He and Bradly C. Stadie and Anne Churchland and Liam Paninski and Yizi Zhang},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.02864},
eprint = {2610.02864},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.02864},
}
RIS
TY - UNPB
TI - NeuroLens: Learning Latent Embeddings of Neural Semantics from Chronic Recordings
AU - Hanrui Lyu
AU - Baiyuan Chen
AU - Tianshu Tan
AU - Matthew R. Whiteway
AU - Maxwell D. Melin
AU - Ji Xia
AU - Linyang He
AU - Bradly C. Stadie
AU - Anne Churchland
AU - Liam Paninski
AU - Yizi Zhang
PY - 2026
DA - 2026/10/02
PB - arXiv
DO - 10.48550/arxiv.2610.02864
UR - https://arxiv.org/abs/2610.02864
AB - Understanding how neural activity represents higher-order cognition and how these representations evolve over time has long been a central pursuit in neuroscience. However, current analytical tools cannot easily distinguish representational plasticity from recording instability in chronic neural recordings. Here, we introduce NeuroLens (Latent Embeddings of Neural Semantics), a self-supervised model based on the Joint-Embedding Predictive Architecture (JEPA) framework that learns denoised, semantically informative latents from chronic neural recordings. An adaptive encoder maps changing neural populations into a common latent space, while a temporal predictor learns structure that supports prediction of future latent states. By predicting in latent space, NeuroLens captures temporally predictable structure and reduces sensitivity to transient, recording-specific variability. Across chronic intracortical data in mice and humans, the learned representations improve decoding of decision-making and semantic task variables. Multi-day pretraining enables generalization to future sessions, rapid few-shot adaptation to unseen neural populations, and more stable decoding over time than state-of-the-art baselines. Together, these results establish NeuroLens as a new paradigm for studying how neural representations change during learning and over long timescales.
ER -