Resources

Cosmic Emulators
Machine Learning resources:

Online courses:

A classic starting point is Andrew Ng’s course (although the assignments are expected in Matlab/Octave): https://www.coursera.org/learn/machine-learning and http://cs229.stanford.edu/

Github repos from workshops:

Specifically for Neural networks/Deep learning:

Python frameworks:
  • scikit-learn: python library that provides a range of machine learning tools – not only supervised and unsupervised learning methods, but also for pre-processing pipelines, validation etc. Not intended for deep learning framework. No GPU support.
  • TensorFlow/Theano: popular python libraries for deep learning. Both have GPU support. Theano is one of the oldest libraries for deep learning. TensorFlow is newer, faster, and has a guaranteed support from Google.
  • Keras (https://keras.io/): a user-friendly python wrapper around Theano and Tensorflow – it has a modular structure and a ‘pythonic’ interface. Recommended for fast prototyping deep networks. But for higher functionalities and advanced computations, one may have to use TensorFlow/Theano.
Blogs regarding algorithms and implementation:

 
SciDAC-4 Home