This repository contains the code and report for the EPFL NX-414 (Brain-like Computation and Intelligence) course mini-project.
This mini-project aims to predict neural responses in the inferior temporal (IT) cortex of non-human primates to visual stimuli. Using the Majaj et al. (Journal of Neuroscience, 2015) dataset, which pairs object images with multiunit neuron activity across IT sites, we compare model strategies from basic pixel-level regression to transfer learning from complex task-driven deep networks and end-to-end data-driven architectures. By evaluating each model’s predictive performance, we seek to identify the framework that best captures IT cortex neural dynamics.
The code and environment.yml are provided for reproducibility. See the report for details about results and analysis.