Few-shot Algorithms for Consistent Neural decoding (FALCON) Benchmark published at NeurIPS 2024

At NeurIPS 2024, Brianna Karpowicz and Joel Ye presented the new Few-shot Algorithms for Consistent Neural decoding (FALCON) Benchmark. FALCON aims to provide rigorous selection criteria for robust intracortical brain-computer interface (iBCI) decoders, easing their translation to real-world devices. The project evaluates decoding algorithms based on their stability for held-out neural data (unseen during decoder training), with the goal of developing algorithms that achieve stable iBCI decoding over weeks, with minimal data collected on new days. The project includes diverse datasets (five curated datasets that span movement and communication iBCI tasks in humans and primates). With the help of AE Studio, we developed a flexible, cloud-based evaluation platform that tests approaches on private neural data.

See the FALCON project page for more info. Congrats Brianna & Joel!!