Rui Ponte Costa

This paper shows how hunger reconfigures a reward circuit in Drosophila to support memory formation by switching dopamine neuron activity modes.

In mice learning two tasks, cerebellar granule cells kept the low-rank geometry of premotor cortical output within each task but rotated their population trajectories apart between tasks, while cortex generalized.

This paper develops a computational framework that unifies reward-based and error-based learning by mapping them into a shared action-gradient space, linking them to brain systems like the basal ganglia and cerebellum.

This paper introduces a theory of hierarchical credit assignment in the cortex, linking cell-type-specific dendritic excitation-inhibition balance to learning and behavior.

This study uses deep reinforcement learning to show that hippocampal circuitry enables learning in partially observable environments, aligning with animal behavior and experimental data.

This paper introduces a computational model of the neocortex that uses self-supervised learning to predict sensory inputs, showing layer-specific predictions and matching experimental observations.

This paper introduces a generalized BTSP rule for neural network learning and shows it can perform unsupervised and supervised tasks but faces challenges with network depth.

This study identifies distinct roles of cortical layer 5 IT and ET neurons in associative learning using transgenic mice and calcium imaging.

This study introduces a computational model showing cerebellar feedback enables rapid task acquisition, switching, and consolidation through cortical dynamics.

The paper investigates how cortical circuits enable region-specific credit assignment in reinforcement learning by analyzing the role of VIP and SST interneurons and cholinergic signals.

This study shows that prefrontal cortex representations change from high-dimensional and nonlinear to low-dimensional and rule-selective as animals learn a new task rule.

This paper introduces a computational model of cerebro-cerebellar interactions that decouples learning from feedback, showing improved performance in motor and cognitive tasks.

The paper models how a developmental shift from synaptic depression to facilitation stabilizes neural activity and controls excitation-inhibition balance.

The paper investigates how reward information and task variables are encoded in midbrain dopamine neurons projecting to different striatal regions, showing heterogeneous and distributed coding.

This paper investigates how pre- and postsynaptically expressed spike-timing-dependent plasticity (STDP) differentially affect neuronal learning in minimalist models.

The paper models spike-timing-dependent plasticity (STDP) expressed pre- or postsynaptically to show how different implementations affect learning outcomes in neuronal networks.

This review discusses the functional consequences of pre- and postsynaptic expression of homeostatic and Hebbian synaptic plasticity, highlighting predicted benefits like improved sensory perception and memory savings.