13-15

Welcome to Machine Learning for Neuroscience!

* Resources:

Here we have compiled some useful machine-learning resources:

Coding resources:

• Scikit-learn: a VERY complete machine learning Python library with amazing documentation and examples.
• Scipy: an also VERY complete math Python toolbox. Includes functions for doing statistics, optimization, signal processing, linear algebra, image processing, etc.
• Nilearn: a Python library specialized to deal with Neuroimaging data!

Papers:

• Domingos, 2012: A Few Useful Things To Know About Machine Learning
• Bzdok et al., 2018: Statistics versus machine learning
• Bzdok et al, 2018: Prediction and inference diverge in biomedicine: Simulations and real-world data
• Bouthilier et al.,2021: Accounting for variance in machine learning benchmarks
• Benkarim et al., 2021: The Cost of Untracked Diversity in Brain-Imaging Prediction
• Vu et al., 2018: A Shared Vision for Machine Learning in Neuroscience
• Josse et al., 2020: On the consistency of supervised learning with missing values
• ... more to come! Stay tune ;)

* Hands-on tutorial requirements:

Using Google Collab

If you do not want to install anything locally, you can use Google Collab to run the demo. To do so you will need a Google account.

1. Download the MAIN_BASC064_subsamp_features.npz and the participants.tsv files found in this repo, and place them into your main Google Drive folder. That's it, you're all set!
2. To open and run the notebook just click

Else, if you do not want to get your hands dirty, and just want to passively follow the demo, you can visualize the results by clicking

Using Python locally

Alternatively, you can also run things locally in your own computer. This requires some extra installation steps:

1. Install Anaconda

2. Create a Conda virtual environment. To do so, open the terminal and type:

   conda create --name ipn_summer_school python=3.6 numpy scipy pandas scikit-learn nilearn seaborn matplotlib
   conda activate ipn_summer_school
   

3. Install Jupyter and add your new conda environment to the kernels available:

   conda install jupyter 
   conda install -c anaconda ipykernel
   python -m ipykernel install --user --name=machine-learning
   

4. Install Nilearn:

   pip install nilearn
   

5. Download the demo.ipynb, MAIN_BASC064_subsamp_features.npz and participants.tsv files found in this repo, and place them into a new directory in your local computer.

6. In the terminal, navigate to the directory where you placed the files:

   cd /path/to/directory
   

7. Open Jupyter-notebook by typing in the terminal:

   jupyter-notebook
   

8. Click on the demo.ipynb file.

You’re ready to go!

NOTE: the name ipn_summer_school can be replaced by anything you want, hopefully a short and meaningful name that tells you what the environment is about. This same environment will be used for the hands-on dynamical systems tutorial.