Machine learning tutorial

ESAC Data Analysis & Statistics Workshop 2017

By Michelle Lochner

To get a copy of this repository type git clone https://github.com/MichelleLochner/ml_esac.git in the command line or click "Clone or download" and click "download zip" if you don't have git installed.

Key files:

machine_learning_notes.pdf -> The notes from the lecture (without the answers)
pre_workshop_questions.pdf -> A set of questions for you to spend ~30 minutes investigating that I'll be asking throughout the lecture
supernova_tutorial.ipynb -> A Jupyter Notebook tutorial for supernova classification with machine learning.

Running the code

Using Anaconda

I strongly recommend using anaconda to run the tutorial code:

Install anaconda if you don't already have it (https://www.continuum.io/downloads)
Create a new anaconda environment by typing (inside the ml_esac folder):

conda env create --name ml --file environment.yml

Activate the environment by typing:

source activate ml

Note: If you have tsch instead of bash this will not work!

A simple workaround is to manually edit your PATH environment variable to point to the new anaconda environment:

setenv PATH <your path to anaconda>/envs/snmachine/bin/:$PATH

Setting up dependencies yourself

If you don't want to use anaconda or create a separate environment, the requirements to run this tutorial code are

dependencies:

python>=3
astropy>=1.1.2
jupyter>=1.0.0
matplotlib>=1.5.1
numpy>=1.11.0
scikit-learn>=0.18.1
scipy>=0.17.0
iminuit>=0.12
sncosmo>=1.3.0

The notebook has not been tested with python 2 but should still work.

Running the tutorial

Type jupyter notebook supernova_tutorial.ipynb into the command line after activating the environment.

Deep Learning tutorial

There is also a very basic deep learning tutorial based on tflearn. You'll need to install tflearn yourself (pip install tflearn). WARNING: this thing is very slow unless you have a GPU and install the GPU version tensorflow. It's also highly incomplete so use at your own risk...