Alex Roxin

The paper shows that random connectivity in neural networks preserves representational similarity despite activity drift, applicable to biological circuits.

The paper shows that locally balanced inhibition prevents synaptic weight over-strengthening in feedforward networks with Hebbian plasticity, maintaining robust learning when input-output correlations are weak.

The paper shows that Hebbian learning explains reduced representational drift in hippocampal CA1 place cells after novel environment exposure by allocating synaptic resources to salient experiences.

The paper shows that representational drift is an inevitable result of ongoing memory storage in neuronal circuits.

The paper identifies a two-phase neural code in hippocampal CA1 pyramidal neurons that encodes time and distance, linking spatiotemporal integration to predictive behavior.

The paper develops a simplified computational model of Behavioral Time-scale Plasticity (BTSP) to study memory storage in recurrent neural networks and shows it enhances memory capacity.

The paper shows that representational drift in hippocampal CA1 pyramidal cells is due to synaptic turnover and Hebbian plasticity, with two distinct time-scales of drift.

The paper investigates how time-varying inputs shape synaptic plasticity and neuronal circuit connectivity, showing that oscillatory inputs with specific phase relationships lead to hierarchical clustering in large networks.

This paper shows that continuous attractor networks can optimally integrate circular stimuli and track their average through activity bump phase, with temporal weighting influenced by internal dynamics and sensory inputs.

This paper shows that an attractor model of perceptual decision making can operate in different integration regimes, including a novel 'flexible categorization' regime with a non-monotonic psychometric curve.

This paper extends meanfield theory to spiking networks with arbitrary degree distributions and correlations, showing they can generate sustained responses to stimuli.

This paper extends the drift-diffusion model to multiple-alternative forced-choice decision making by deriving a framework that accounts for nonlinear dynamics and correlated noise.

The paper investigates how oscillatory inputs affect network dynamics in cognitive tasks, finding that slow oscillations activate memory recall states while faster oscillations clear memory states.

The paper shows how theta-modulation accelerates learning by enabling the formation of environment-specific synaptic connectivity patterns in hippocampal circuits during spatial exploration.

The paper analyzes cortical dynamics in anesthetized rats, finding non-rhythmic UP-DOWN state transitions in a bistable balanced network.

The paper shows that small sample sizes make it hard to distinguish different network topologies in cortical circuits and introduces a new statistic to overcome this limitation.

This paper derives mean-field equations for neuronal networks with arbitrary degree distributions and shows they exhibit richer dynamics than Erdos-Renyi networks.