My laboratory pursues theoretical and computational research on the neural basis of sensory and cognitive processes. These include memory, decision making, and more recently the chemosensory processes involving taste perception. Priority is given to developing biologically plausible models, often in terms of populations of spiking neurons, using techniques borrowed from physics, machine learning, and data science. With this approach we have characterized the metastable nature of ongoing and evoked activity in several cortical areas (most notably the primary gustatory cortex), and worked out some of its consequences for neural coding. We hope that uncovering the common basis of ongoing and evoked activity can tell us much about how cortical networks are organized and function. The investigation of what type of sensory and cognitive functions can be subserved by metastable (rather than stable) neural activity is also a central effort of the lab.

We are also interested in the theory of learning and modeling synaptic plasticity in cortical circuits. On this topic we have made progress on two main fronts: 1) We have proposed a model of how to segment a sensory stream to extract the patterns of neural activity of unknown nature and timing that are relevant  for making decisions. This was achieved by reinforcement learning in a network of spiking neurons, and is a first step towards understanding how an agent can learn to identify the 'features' of its environment that are relevant for decision making. 2) We have proposed a biologically plausible plasticity rule that tunes a spiking network into the metastable dynamical regime we have observed and analyzed in depth in cortical data. The plasticity co-exists with ongoing metastable dynamics (the bulk of the synaptic weights are stable against ongoing neural activity) and can learn new representations by training with new stimuli (the weights remain plastic), offering a potential solution to the long-standing stability-plasticity dilemma. 

While the main focus of our lab is the mammalian brain (especially the neocortex), our interests remain eclectic and we are always willing to take on intriguing challenges. For example, we have recently proposed the first mathematical model of decentralized vision in the sea urchin, an invertebrate that orients to visual stimuli despite lacking eyes.