Resources and Code

A fast, modular, and language-agnostic platform designed to integrate artificial neural networks into real-time neuroscience experiments, enabling parallel data acquisition, control, and analysis with low latency.

A benchmark for comparing dynamics in task- and data-trained models.

A benchmark to standardize the evaluation of adaptation algorithms for intracortical brain-computer interfaces.

A deep learning method designed to infer latent dynamics from single-trial neural spiking data.

A Transformer-based method of modeling neural spiking data without an explicit dynamics model.

A benchmark suite introduced to standardize latent variable modeling of neural population activity, facilitating the analysis of diverse neural systems and behaviors and promoting the use of unsupervised evaluation across various datasets.

A deep learning method for enhancing stability of intracortical brain-computer interfaces by updating the mapping of nonstationary neural data, enabling accurate behavioral decoding over extended periods without supervised recalibration.

A method combining injective readouts with latent dynamics models to provide accurate and interpretable reconstruction of neural data with few latent dimensions.

A deep learning method for enhancing the estimation of network state and dynamics from two-photon calcium imaging.