Mehdi Khamassi

The paper investigates how the brain encodes epistemic value during goal-directed causal learning, showing that expected information gain is represented in sensorimotor regions and challenges traditional views of reward-based valuation.

The paper compares curiosity models using experimental results and mathematical formulations to understand how motivation and exploration evolve in uncertain environments.

This paper introduces the Epistemic Replay Algorithm (ERA) to explain hippocampal replay by integrating curiosity and reward, showing it better accounts for replay properties than non-curious models.

This paper introduces a reinforcement learning agent that adapts to changing environments by detecting change points retrospectively and managing multiple local models.

This study shows that humans adapt to rapid environmental changes by inferring strategies from an abstract space, not just through incremental learning.

The paper examines how ACC neurons in rats respond to rule or strategy changes during learning tasks and shows that hippocampal theta and sharp-wave ripples modulate these responses differently.

The paper shows that increasing cortical plasticity via RGS14414 overexpression improves one-trial memory but causes memory interference and alters hippocampal-cortical interactions.

The paper identifies a brain region in Rhesus macaques that shows computational properties similar to the human temporo-parietal junction, modulated by social interaction predictability.

The paper uses the Object Space Task to show Ehmt1 +/- mice have reduced episodic-like memory and exploration but enhanced semantic-like memory, with semantic features better for genotype classification.

This study shows that dopamine modulates the exploration-exploitation trade-off in rats by increasing exploration when dopaminergic activity is reduced.

This paper investigates strategic learning in social interactions, showing that sophisticated belief-based learning leads to more accurate beliefs and convergence toward game optimality, influenced by game structure and individual cognitive capacity.

This paper bridges computational neuroscience and machine learning by combining bio-inspired meta-learning, Kalman filter, and exploration bonuses in non-stationary multi-armed bandit tasks, showing improved performance over existing methods.