Related Experiment Video
Updated: May 13, 2026
11:18
Closed-loop Neuro-robotic Experiments to Test Computational Properties of Neuronal Networks
Published on: March 2, 2015
11.5K
Bidirectional cross-day alignment of neural spikes and behavior using a hybrid SNN-ANN algorithm.
Binjie Hong1,2, Zihang Xu1,2, Tengyu Zhang3
1Center for Excellence in Brain Science and Intelligence Technology, State Key Laboratory of Brain Cognition and Brain-inspired Intelligence Technology, Institute of Neuroscience, Chinese Academy of Sciences, Shanghai, China.
Frontiers in Neuroscience
|February 20, 2026
Summary
AlignNet enhances neural decoding by aligning brain signals and behavior across days. This novel framework improves cross-day neural interface applications by learning session-invariant representations.
Area of Science:
- Neuroscience
- Deep Learning
- Computational Neuroscience
Background:
- Deep learning effectively interprets electroencephalogram (EEG) signals.
- Invasive brain signals face challenges in cross-day neural decoding and simulation due to non-stationarity and representational drift.
Purpose of the Study:
- To present AlignNet, a novel framework for cross-modal alignment between neural spiking patterns and behavioral semantics.
- To enable robust cross-day neural decoding and simulation by addressing representational drift.
Main Methods:
- AlignNet utilizes hybrid Spiking Neural Network-Artificial Neural Network (SNN-ANN) autoencoders for representation learning.
- A U-based architecture encodes neural spikes and behavior into a shared latent space, optimized via contrastive objectives for session-invariant features.
- A pretraining strategy using multi-session data is employed, followed by task-specific fine-tuning.
Main Results:
- AlignNet demonstrates superior performance in both single-day and cross-day neural decoding and simulation tasks.
- The pretrained AlignNet model effectively performs decoding and simulation after fine-tuning.
- Hybrid SNN-ANN representations show temporal consistency across multi-day recordings while preserving behavioral semantics.
Conclusions:
- AlignNet offers a robust solution for cross-day neural decoding and simulation.
- The framework advances the development of reliable and adaptable cross-day neural interface applications.
- Session-invariant feature learning is crucial for generalizing neural decoding across different recording sessions.
