Machine Learning & Deep Learning for Neuroscience

A 12-week instructor-guided course on the NeuroAI track, running from October 25 2026 to January 10, 2027. It builds systematically from machine-learning foundations through neural networks and deep learning for neural data, and closes with representation learning and a first look at modern architectures.

The course is organised into five modules: Machine Learning Foundations (ML paradigms, regression and classification, classical ML algorithms, framing neuroscience problems), Unsupervised Learning and Reliable Evaluation (clustering, PCA and dimensionality reduction, validation and generalization, data leakage and reproducible pipelines), Neural Networks and Learning (perceptrons and multilayer networks, activations and loss functions, gradient descent and backpropagation, Hebbian learning and STDP, biological plausibility), Deep Learning for Neural Data (CNNs for spatial and image-like data, RNNs / LSTMs / GRUs for temporal data, neuroscience coding tutorials), and Representation Learning and Modern Architectures (autoencoders and VAEs, LFADS as a neuroscience example, introductory exposure to transformers and GNNs, self-supervised and generative learning, final project).

Each week combines about 30–45 minutes of recorded preparatory material, a 60–75 minute live conceptual class, a 45–60 minute guided tutorial, a structured practice notebook, a short quiz or submission, and discussion with instructor support through the Neurovidya course space. Learners do not merely hear about methods — they repeatedly apply them to neuroscience data and receive feedback.