Vikash Gilja

Probabilistic Co-Control in Brain-Computer Interfaces: Uncertainty as a Control Signal in Brain-to-Text Decoding

This paper improves uncertainty estimation in brain-to-text BCIs by modifying training objectives and evaluation metrics, leading to more reliable and informative predictions.

Compositional decoding of neural activity enhances generalization in handwriting BCIs

This paper introduces a compositional decoding model for handwriting BCIs that improves generalization by leveraging behavioral motifs.

SONIC: A Benchmarking Paradigm for Brain-Computer Interfaces

This paper introduces SONIC, a benchmarking paradigm for BCIs, demonstrating an ITR of over 200 bps using an ovine model and the Connexus BCI.

Few-shot Algorithms for Consistent Neural Decoding (FALCON) Benchmark

The paper introduces FALCON, a benchmark suite for evaluating robustness of iBCI decoders using standardized datasets and evaluation metrics.

Cellular Calcium Activity at Depth Predicted from Surface Potential Recordings using Ultra-high Density Transparent Graphene Arrays

The paper develops high-density transparent graphene electrodes to record neural activity from deeper cortical layers and decodes calcium activity from surface potentials using dimensionality reduction and neural networks.

Imaging through Windansee electrode arrays reveals a small fraction of local neurons following surface MUA

The paper uses two-photon calcium imaging and surface electrophysiology to show that only a small fraction of layer 2 neurons follow surface MUA, suggesting MUA may primarily reflect axonal inputs.

Local Field Potentials in a Pre-motor Region Predict Learned Vocal Sequences

The paper uses local field potentials in zebra finches to decode vocal sequence elements and shows similarities to human motor areas.

Microscale physiological events on the human cortical surface

The paper identifies three distinct classes of cortical activity using high-density microelectrodes, showing they are modulated by stimuli and present across species.

The Argo: A 65,536 channel recording system for high density neural recording in vivo

The paper introduces the Argo System, a high-density neural recording system with 65,536 channels, validated in rats and sheep for cortical recordings.

Sequence Transfer Learning for Neural Decoding

This paper demonstrates the use of LSTM networks for neural decoding in BCIs, showing improved accuracy and robustness compared to traditional linear models.

A neural decoder for learned vocal behavior

This paper introduces a neural decoder for birdsong, a complex learned vocal behavior, using Brain Machine Interfaces.