Learning Time-Scale Invariant Population-Level Neural Representations

2025-11-17

AI Summary

Current neural time-series foundation models suffer from limited generalization due to inconsistent temporal-scale preprocessing between pretraining and downstream tasks, resulting in representations lacking scale invariance. To address this, we propose Time-Scale Augmented Pretraining (TSAP), the first method to explicitly model and enhance time-scale robustness in population-level neural representation learning. TSAP employs multi-scale temporal transformation-based data augmentation and a temporally encoded architecture that jointly models population-wide channels, enabling cross-scale feature alignment. Experiments demonstrate that TSAP significantly improves robustness to temporal scaling and sampling-rate variations in downstream decoding tasks—particularly in brain–computer interfaces—while enhancing representation transferability and practical utility. This work establishes a novel paradigm for developing general-purpose neural time-series foundation models.

Technology Category

• Machine Learning: Time-Series/Data Streams
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Planning, Routing, and Scheduling: Temporal Planning

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

General-purpose foundation models for neural time series can help accelerate neuroscientific discoveries and enable applications such as brain computer interfaces (BCIs). A key component in scaling these models is population-level representation learning, which leverages information across channels to capture spatial as well as temporal structure. Population-level approaches have recently shown that such representations can be both efficient to learn on top of pretrained temporal encoders and produce useful representations for decoding a variety of downstream tasks. However, these models remain sensitive to mismatches in preprocessing, particularly on time-scales, between pretraining and downstream settings. We systematically examine how time-scale mismatches affects generalization and find that existing representations lack invariance. To address this, we introduce Time-scale Augmented Pretraining (TSAP), which consistently improves robustness to different time-scales across decoding tasks and builds invariance in the representation space. These results highlight handling preprocessing diversity as a key step toward building generalizable neural foundation models.

Problem

• Addressing sensitivity to time-scale mismatches in neural population representations
• Improving robustness across preprocessing variations in neural decoding tasks
• Building invariant representations for generalizable neural foundation models

Innovation

• Time-scale Augmented Pretraining improves robustness
• TSAP builds invariance in representation space
• Addresses preprocessing diversity for neural foundation models