Leo Kozachkov

This paper introduces DSA, a metric to compare systems by their dynamics, showing it captures dynamical structure missed by geometric and topological methods.

This study uses DeLASE to show that propofol anesthesia reduces dynamic stability in macaque cortex, mirroring destabilized linear systems.

This study shows propofol disrupts sensory signal progression through the cortical hierarchy by affecting both Up and Down states, with persistent responses only in auditory cortex.

The paper shows that neuron-astrocyte networks can perform the core computation of a Transformer AI architecture, providing a biological basis for its success across multiple domains.

This study shows that short-term synaptic plasticity (STSP) improves the robustness and brain-like behavior of artificial recurrent neural networks in working memory tasks.

The paper uses control theory to analyze stability mechanisms in recurrent neural networks, identifying synaptic plasticity and balance as key factors for stable dynamics.

The paper uses control theory to analyze stability mechanisms in neural networks, identifying anti-Hebbian plasticity and synaptic sparsity as key factors for maintaining stable dynamics.