NeuroAI and Machine Learning in Neuroscience

What it is about

You will learn modern methods of Computational Neuroscience and NeuroAI: from model-based approaches (dendrite models, mean-field, Fokker-Planck, linear response) through plasticity and stochastic processes to training methods for RNNs and SNNs as well as Reinforcement Learning and Predictive Coding. In the end you will be able to implement models at different levels, train RNNs/SNNs and analyze and reproduce NeuroAI papers in a professional manner.

What you will be able to do

• Understand advanced topics: mean-field models, stochastic processes, efficient coding, Reinforcement Learning, training of RNNs
• Compare the challenges of deeper learning methods with the brain's sparse, event-driven processing
• Describe models of memory formation and biologically plausible backpropagation
• Analyze and train various recurrent networks (e.g., Reservoir Computing, e-prop)
• Understand principles of Predictive Coding and its role in biological networks
• Explain the use of machine‑learning tools (e.g., simulation-based inference) for model inference from neural data
• Implement and evaluate network models at different levels (mean-field, rate-based, spike-based)
• Decompose NeuroAI publications into the components covered in the course

What the module consists of

• LectureDelivery of theoretical concepts in NeuroAI and Machine Learning for Neuroscience
• Tutorials / Peer‑ProgrammingConsolidation of lecture material through interactive notebooks and programming tasks
• ProjectApplication of concepts to reproduce a scientific paper with final presentation

Teaching method

• Lecture (PowerPoint)Introduction and explanation of theoretical foundations
• Hands‑on Tutorials with interactive notebooks and peer‑programmingPractically practicing implementations and methods
• Project with independent and supervised sessionsApplying and deepening what is learned through reproduction work and presentation