Neural Coding and Chaos in Recurrent Networks
example code: github.com/glajoie/torus_flow_spiking_network/
I am interested in neural coding, or the ability of neural networks to encode information from a stimulus that perturbs its dynamics. I investigate the reliability of excitable neural networks, driven by a given input signal. Reliability can be described as the ability of a dynamical system to reproduce the same output, given a single stimulus, on many trials where initial conditions change. Moreover, I wish to address questions on information carrying capacities of such networks and the role of reliable behavior in this context: When are excitable neural networks reliable? What are the implications for possible encoding schemes given a reliable (or unreliable) network? Here, I use a blend of bifurcation theory, numerical simulations and information theoretic tools to attack these questions.
example code: github.com/glajoie/BBCI_spiking_plasticity_model/
I am generally interested in problems involving driven recurrent neural dynamics by by neural implants. Neuronal networks in the brain react to external inputs -in the form of modulated electrical currents- on a variety of timescales, from immediate evoked spiking activity, to changes in population dynamics, to changes in connectivity strengths (synapses) within networks. Some of my earlier work focused on immediate dynamical changes induced by neural implants at the level of population synchrony. This is intimately linked to studies of neural pathologies such as Parkinson's disease and their treatments via neuroprosthetics (Deep Brain Stimulation). More recently, I have been working on theoretical frameworks to understand the effects of Bidirectional Brain-Computer Interfaces (BBCI) on cortical networks. Experiments show that spike-triggered stimulation performed with BBCI's can artificially strengthen synaptic connections between distant Motoc Cortex (MC) sites and even between MC and spinal cord sites, which can produce changes that last several days. Here, a neural implant is triggered by spikes from an MC site and electrically stimulates a secondary target site after a set delay, the value of which is critical in determining the efficacy of the procedure. In parallel with ongoing experiments, I am developing a recurrent network model with probabilistic spiking mechanisms and plastic synapses capable of capturing both neural and synaptic activity statistics relevant to BBCI protocols.
Tradeoff between precision and speed for perception of rapidly changing sensory stimuli
example code: github.com/glajoie/V1_ring_model/
Sensory systems of animals have evolved to optimally process natural stimuli and ensure their survival. Depending on the environment, emphasis can be put on the precision of the perception (e.g.~low-contrast vision to hunt in the dark) and/or on its rapidity (e.g.~detecting the rapid motion of a predator). Given that sensory systems are implemented with populations of interacting neurons, what are the organizational features that influence the relationship between precision and rapidity? Are there foundational principles | either at the level of encoding statistics or population dynamics | that could provide insights into psychophysical attributes of perception and even help design artificial systems? My research explores these questions in the early visual system of primates, but has an outlook toward different sensory modalities.