Josue Ortega Caro
Machine Learning for Neural Dynamics
Wu Tsai Fellow · Yale University · Cardin Lab
Computational Neuroscience
Foundation Models
Generative Models
Inductive Bias
Machine Learning for Neural Dynamics
Wu Tsai Fellow · Yale University · Cardin Lab
Combining deep learning, mathematical modeling, and neuroscience to tackle fundamental questions about brain computation.
BrainLM is a foundation model trained on 6,700 hours of fMRI recordings. It predicts clinical variables, forecasts future brain states, and discovers functional networks through zero-shot inference.
ICLR 2024
Attentional Neural Integral Equations (ANIE) learn unknown integral operators from data, providing a principled framework for modeling spatiotemporal dynamics in physical and biological systems.
Nature Machine Intelligence 2024
A structured multimodal transformer combining masked autoencoding with causal attention to reveal selective changes in cholinergic signaling during visual perceptual learning.
bioRxiv 2025
Benchmark platform for evaluating whether interpretability methods recover latent computation, mechanism class, support structure, and intervention effects in synthetic neural dynamics.
Benchmark Platform
Robust deep learning models rely on low-frequency information. Local convolutions induce an implicit bias toward high-frequency adversarial examples via the Fourier Uncertainty Principle.
PLOS Comp. Bio. 2023 · Frontiers 2024