Recent advances in machine learning methods for computational chemistry have opened exciting new avenues for understanding olfaction. We are utilizing graph neural networks and self-supervised learning to identify patterns in unlabeled molecular data, generating representations of odors that are useful for predicting olfactory sensory neuron activation. These learned representations more accurately predict how individual odor molecules interact with olfactory receptors compared to existing methods.

Relevant poster: Improved odor-receptor interaction predictions via self supervised learning
McConachie, G, Duniec, E, Younger, M, DePasquale (2024)
NAIsys CSHL 2024

State-space models (SSMs) are powerful tools for modeling time series data that naturally arise in neuroscience, finance, and engineering. These models assume observations arise from a hidden latent sequence, encompassing methods like Hidden Markov Models (HMMs) and Linear Dynamical Systems (LDS). Our group is actively building an open-course software packge to implement these models in Julia called StateSpaceDynamics.jl. This modular package designed to be fast, readable, and self contained for the express purpose of fitting a plurality of SSMs, easily in Julia.

Relevant poster: StateSpaceDynamics.jl: A Julia package for probabilistic state space models (SSMs)
Senne, R., Loschinskey, Z., Loughridge, C., Fourie, J., DePasquale, B. (2024)
in prep

Biological neural networks compute differently than most artificial neural networks used in machine learning. For example, although real neurons communicate with spikes, reproducing this feature in artificial models has been a challenge. We develop methods for training biophysically detailed neural networks and use thse models to understand how real biologial circuits compute. Through mathematical modeling, we focus on building tighter links between biologial neural networks and more abstract artifical neural network models used in machine learning.

Relevant paper: The centrality of population-level factors to network computation is demonstrated by a versatile approach for training spiking networks
DePasquale, B, Sussillo, D., Abbott, L.F., Churchland, M.M. (2023)
in press at Neuron

Neural recordings from behaving animals are often much too complex to link directly to an animal’s ongoing behavior. We develop machine learning models to analyze complex neural datasets to understand the algorithms that underlie different behaviors. We principally focus on the neural underpinnings of movement and decision making.

Relevant paper: Neural population dynamics underlying evidence accumulation in multiple rat brain regions
DePasquale, B., Brody, C.D., Pillow, J. (2022)
in revision at eLife