Course ML at NERSC

Internal ML course/practical demonstration intern to NERSC

Link to the slides

• Session 1: Introduction, generalities on machine learning (pdf)
• Session 2: Validation, overfitting, regularization (pdf)
• Session 3: Random Forest, grid search (pdf)
• Session 4: Neural networks (pdf)
• Session 5: Convolutional neural networks (pdf)

Practical demonstration

Instruction for working on a cloud (recommended)

Run the tutorial in a cloud computing provider (require Google login):

• Practice 1: Introduction and linear regression
• Practice 2: Validation, overfitting, regularization
• Practice 3: Random forests. Grid search.
• Practice 4: Neural networks.
• Practice 5: Convolutional Neural Networks and Regularizations.
• HACKATHON Data for hackathon.

Instructions for working locally

You can also run this notebook on your own (Linux/Windows/Mac) computer. This is a bit snappier than running them online.

1. Prerequisite: Python>=3.7. If you're not a python expert: 1a. Install Python via Anaconda. 1b. Use the Anaconda terminal to run the commands below. 1c. (Optional) Create & activate a new Python environment. If the installation (below) fails, try doing step 1c first.

2. Install: Run these commands in the terminal (excluding the $ sign): $ git clone https://github.com/nansencenter/nersc_ml_course.git $ pip install -r nersc_ml_course/requirements.txt

3. Launch the Jupyter notebooks: $ jupyter-notebook This will open up a page in your web browser that is a file navigator.
   Enter the folder nersc_ml_course/notebooks, and click on the tutorial notebook_name.ipynb