Machine Learning



The Machine Learning focus group will span over PWG, hardware and computing activities that aim at developing and delivering machine learning techniques for STAR. As computational related, the pages starts under the S&C Drupal

Our main goal and is to see analysis techniques using ML to be presented and shared, discussed and improved as well as lesson learn and findings from analyzers passed from one to another for a mutually beneficial outcome.  ML techniques are especially hard to optimize and your shared experience would be an asset to others. Similarly, we hope to see the development of GANs and other techniques that may propel the Physics capabilities of STAR (speed of GANs over standard simulator would be a huge step forward for STAR).

Jerome Lauret - S&C Leader 2018

ML Tools and Tutorials :

Tools - 

  1. Scikit-Learn - 
    Easiest to start and play around with - comes standard and integrates with your existing python installation
  2. Keras -
    State of the art libraries and functions with predefined networks and relatively easy to implement examples
  3. Tensorflow -
    Google product and most often used as the backend for keras libraries to run on. You can also have tensorflow specific code as you follow up from the example
  4. Theano -
    Also a backend for keras. main difference w/ tensorflow are in how images are defined (x, y, pixel-w) vs (pixel-w, x, y). its being less used now
  5. Cuda toolkit -
    A bit of an advanced toolkit for more serious developers. fully contained set of libraries in c++ code, takes advantage of gpu optimization
  6. PyTorch -
    front end with native python setup similar to Keras
  7. Mxnet -
    more of a backend alternative to tensorflow

Tutorials -

  1. Building your first network : Im very fond of this website! it is one of the best in explaining the details and proving hands on examples for you to learn and play. 
  2. Learning from data - 
  3. Stanford classes about CNNs and vision
  4. Another class about vision and object recognition
  5. CMU course on deep learning
Podcasts -

  1. Our own Nick Elsey has a podcast about this exact topic. How did AI algorithms come about, what they are
  2. There is also the NVIDIA podcast about AI which is quite good.


Textbooks -

1. PRML by Bishop 
2. Machine Learning: A probabilistic perspective by Murphy
3. Convex Optimization by Boyd et. al.
4. Deep Learning by Goodfellow et. al.

Useful Seminars/Workshops/Conferences :

  1. Machine Learning For Jets (ML4Jets) FNAL Nov 14-16th 2018 There is vidyo option for those interested to follow remotely (no registration fee!)
  2. Machine Learning Seminar  EIC Center in Jefferson Lab Nov 6th 2018 - 11am EST  and Blue Jeans 373678588
  3. 2nd IML Workshop @ CERN April 2018
    This is part of the Inter Experimental LHC Machine Learning Working Group (IML)
    CERN They host many excellent seminars and workshops. We will follow it and repost the ones of interest here
  4. Machine Learning for Jet Physics (ML4Jets) LBL Dec 2017  Mostly jet oriented applications
  5. Data Science @ HEP FNAL May 2017  Excellent set of slides and tutorials over a range of topics
  6. Data Science @ LHC CERN 2015  - useful slides and tutorials

For any questions/comments/suggestions please contact
Raghav Kunnawalkam Elayavalli ( & Daniel Brandenburg (