We investigate the thermodynamics of neural computation — specifically, how biological networks might exploit noise-assisted transport to solve the energy–information bottleneck of hypothesis selection.
Our primary framework is Coherent Resonant Netting (CRN), which models a two-regime decision architecture: a low-cost wave-like filtering stage (Stage-I) that prunes hypotheses before expensive spiking commitment (Stage-II). Using GKSL/Lindblad open-system dynamics as a functional proxy, we test whether Disorder-Enhanced Selectivity (DES) emerges on real biological connectomes and depends on native network topology.
Current work spans 100 Human Connectome Project subjects and three model organisms (C. elegans, Drosophila larva, mouse cortex proxy), with over 500,000 transport simulations and full topology-destroying controls. All code and data are open.
Background in applied optimization and distributed systems (20+ years). Since 2020, focused exclusively on theoretical neuroscience: Landauer bounds in biological networks, spectral graph theory, ENAQT regimes, and variational free energy minimization.