Embodied Neurocomputation: A Framework for Interfacing Biological Neural Cultures with Scaled Task-Driven Validation
This research proposed an Embodied Neurocomputation framework to interface biological neural networks (BNNs) with silicon computing for navigation tasks. By systematically optimizing encoding parameters, the study found specific settings that allowed biological agents to learn robustly over many episodes, outperforming optimized silicon Deep Q-Networks under the same training budget.