Searching for low-dimensional neural manifolds

Speaker: Dr. Michele Allegra (Padova Neuroscience Center, “Galileo Galilei” Physics and Astronomy Department)

When: 25 January 2024, h. 5:00 pm

Where: 1C150, Torre Archimede

Abstract: Traditionally, the relation between cognition and the brain was understood in terms of neurons performing a sequence of operations on input signals. In recent years, the advent of large-scale neuronal recordings has shifted the focus on movement within representational spaces collectively implemented by neural populations. Geometrically, we can picture these spaces, termed “neural manifolds”, as low-dimensional surfaces in a neural population state space in which each axis represents the activity of one neuron. Modelling movement along the manifold provides a dynamical-system description of neural computation. In this talk, I will provide an introduction to the topic, focusing on how neural manifolds can be characterized and reconstructed from the data by combining geometrical and probabilistic methods.