analiza aiqbench
NeuroLens applies the Joint-Embedding Predictive Architecture to chronic neural recordings, effectively separating true representational plasticity from recording noise in both mice and humans. This self-supervised approach enables stable decoding of semantic variables across days and facilitates rapid adaptation to new neural populations without extensive retraining. The work advances the intersection of machine learning models and neuroscience by providing a robust framework for studying long-timescale cognitive evolution.
NeuroLens: Learning Latent Embeddings of Neural Semantics from Chronic Recordings
arXiv:2610.02864v1 Announce Type: new 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.
Authors:Hanrui Lyu, Baiyuan Chen, Tianshu Tan, Matthew R. Whiteway, Maxwell D. Melin, Ji Xia, Linyang He, Bradly C. Stadie, Anne Churchland, Liam Paninski, Yizi Zhang
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.
Machine Learning (cs.LG); Neurons and Cognition (q-bio.NC)
arXiv:2610.02864 [cs.LG]
(or arXiv:2610.02864v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2610.02864 arXiv-issued DOI via DataCite (pending registration)
Text integral sub licența deschisă a publicației · 220 cuvinte
Subiecte conexe
Cross-Subject Semantic Decoding with Shared-Space Alignment for Generalized Neural Representation Learning
arXiv · 23 iul. 2026
Exploring Hierarchical Interest Representation For Meta Ads Deep Funnel Optimization
Meta Engineering · 15 iul. 2026
LatentFlow: Visual Analytics for Latent Space Analysis in Molecular Graph Neural Networks
arXiv · 27 iul. 2026
A Universal Dense Football Event Representation Based on TabTransformer
arXiv · 9 iun. 2026