N Alex Cayco Gajic is a Junior Professor in the Département d’Etudes Cognitives at the Ecole Normale Supérieure in Paris, France. She completed her PhD in Applied Mathematics at the University of Washington (Seattle, USA) under the supervision of Eric Shea-Brown. Her dissertational work focused on mathematically analysing the coding implications of higher-order statistics in neural population activity using methods from statistical physics, dynamical systems, stochastic modelling, and information theory. After obtaining her PhD, she made the leap across continents and disciplines to the Department of Neuroscience, Physiology, and Pharmacology at University College London (London, UK) for a four-year postdoc with Angus Silver. In the Silver Lab, she studied pattern separation in the cerebellar cortex, while gaining new skills in detailed circuit modelling and quantitative analysis of large-scale calcium imaging data. She started her own team at ENS in 2019 and strives to integrate both currents of her training – her mathematical foundations and cerebellar expertise – in her faculty research, which focuses on understanding the link between associative learning and cerebellar computation.
Our laboratory studies how neural population activity controls motor learning and how this process is shaped by neural circuitry, with a focus on the cerebellum. We use a broad variety of quantitative methods including dynamical systems, statistical modelling, and machine learning. Our research is spanned by three overlapping axes. First, we are interested in understanding how learning of complex tasks is distributed across different learning centers in the brain. To probe this, we build machine learning inspired models to understand how the cerebellar architecture and error-based learning is able to interact with other brain regions that govern motor learning (such as the motor cortex). Second, we work closely with experimentalists to understand the role of cerebellar circuitry in the context of complex, multiscale behaviors. We use both dimensionality reduction methods to understand neural representations during behavior, as well as statistical models to understand mixed processes behind behavioral learning. Finally, a major component of understanding population activity is the use of powerful methods for analyzing neural representations. Therefore a third axis of our work is to develop mathematically interpretable methods for dimensionality reduction specifically to probe how neural geometry and dynamics are shaped over the course of learning. Together we hope these three axes will help to advance our understanding of how distributed learning centers in the brain coordinate during the acquisition of complex, naturalistic behaviors.