Tutorials for each day of BAMB! 2024. Each folder contains two tutorial versions:

• basic tutorial file, not or only partially filled (eg. modeling_101.ipynb)
• tutorial with solutions (modeling_101_solutions.ipynb)

• Introductory lecture: What is a model ?
• Lectures, divided in 3 parts: part A - Model definition and estimation; part B - Parameter fitting and recovery; part C - Model comparison
• Tutorial, also divided in 3 parts: part A ; part B ; part C

• Tutorials:
  • Introduction and structure
  • Part 1: RL Basics:
    • Overview
    • 
  • Part 2: Fitting RL models to behavior:
    • Slides
    • Colab notebooks:
      • Instructions
      • Solutions

• Tutorials

  • Part 1: Introduction and Bayesian Brain

    • Instructions:
    • Solutions:
    • Slides

  • Part 2: Complex behavior and connections

    • Instructions:
    • Solutions:
    • Slides

  • Part 3: Simulation-based inference: Tutorial

• Tutorials

  • Part 1: Gaussian mixture models and EM

    • Instructions:
    • Solutions:
    • Slides
    • Probabilities cheat sheet

  • Part 2: Hidden Markov models

    • Instructions:
    • Solutions:
    • Slides
    • Probabilities cheat sheet

• Tutorials
  • Parts 1-4: Simulating and fitting the DDM
    • Instructions:
    • Solutions:
    • Slides

In the first part of the tutorial, we will learn how to use Perceptron and Multi-Layer Perceptron (MLP) neural networks to solve simple classification tasks. We will walk through the necessary steps to implement and train these networks using PyTorch. Here is the index:

• Imports.
• Models:
  • Perceptron.
  • Multi-layer perceptron.
• Generate data.
• Training and evaluating.
• Playing

In the second part of the tutorial we will learn how to use recurrent neural networks (RNNs) to solve a Perceptual Decision Making task. Here is the index:

• Installs, packages, auxiliary functions.
• Preparing for the training:
  • Training parameters.
  • Define the task (sample dataset).
  • Define the network.
  • Define the algorithm to train the network.
  • Save config.
  • Supervised training of the RNN.
  • Run the trained network (and save the behavioral data).
• Network analysis:
  • Behavioral analysis.
  • General neural analysis.
  • Stimulus and choice decoding from network activity.

• Lecture slides
• Tutorial slides
• Tutorial instructions: Feedforward models
• Tutorial solutions: Feedforward models
• Tutorial instructions: RNNs
• Tutorial solutions: RNNs
• NeuroGym
• PyTorch

Author (tutorials): Manuel Molano