Focus on: 
Back to Conference View
Login

1st Terascale School of Machine Learning

→ Europe/Berlin
SR4 (DESY)

SR4

DESY
Description

The school aims at PhD students and young postdocs, and offers not only a basic introduction to machine learning, but also talks from leading experts in the field on their current research projects. The list of lectures comprises:
- Introduction to Machine Learning
- Introduction to Deep Learning
- Latest developments and new challenges in the field
- Machine learning outside High Energy Physics

In addition to the lectures, several hands-on tutorials will be offered, covering:
- TMVA
- PyTorch
- Keras and Tensorflow

One afternoon is reserved for a ML challenge.

In order to participate successfully in the school, we expect students to have at least basic knowledge of Python and numPy. If you don't have experience with it, please register to our pre-courses on Monday morning. The official school program will start on Monday afternoon. Please bring your own laptop for the tutorials.

The registration fee is 60 Euro and has to be paid in cash during the school. The number of attendees is limited to 80, in order to get a place please register early.

DESY map
Poster
  • Monday, 22 October
    • 1
      Introduction to Python (optional pre-tutorial)
      Link to GitHub
      Slides
    • 11:00
      Coffee break
    • 2
      Introduction to numPy (optional pre-tutorial)
    • 13:00
      Registration
    • 3
      Welcome
      Speaker: Dr Isabell-A. Melzer-Pellmann (DESY)
      Slides
    • 4
      Introduction to Machine Learning (part I)
      Speaker: Fedor Ratnikov
      Slides
    • 15:40
      Coffee break
    • 5
      Introduction to Machine Learning (part II)
    • 6
      TMVA Tutorial
      Speaker: Lorenzo Moneta
      Link to tutorial
      Slides
  • Tuesday, 23 October
    • 7
      Introduction to Deep Learning (part I)
      Speaker: Dr Dirk Kruecker (DESY)
      notes
      Slides
    • 11:00
      Coffee break
    • 8
      Introduction to deep learning + pyTorch tutorial (part II)
      Speaker: Dr Dirk Kruecker (DESY)
    • 13:00
      Lunch break
    • 9
      Introduction to Deep Learning + pyTorch tutorial (part III)
      Speaker: Dr Dirk Kruecker (DESY)
    • 15:30
      Coffee break
    • 10
      Introduction to Deep Learning + pyTorch tutorial (part IV)
      Speaker: Dr Dirk Kruecker (DESY)
  • Wednesday, 24 October
    • 11
      Deep Learning - Convolutional and recurrent networks
      Speaker: Gregor Kasieczka (Institut fuer Experimentalphysik / UHH)
      Slides
    • 11:00
      Coffee break
    • 12
      Adversarial Frameworks - from traditional GANs to WGANs and progressive GANs
      Speaker: Mr Jonas Glombitza (RWTH Aachen)
      Slides
    • 12:50
      Lunch break
    • 13
      Tutorial: Tensorflow
      Speaker: Mr Adam Elwood (DESY)
      Slides
    • 15:30
      Coffee break
    • 14
      Tutorial: Keras
      Speaker: Dr Christian Contreras Campana (DESY)
      Slides
    • 17:50
      Break
    • 15
      Networked data-science for research, academic communities and beyond
      There is an exceptional way of doing data-driven research employing networked community. The following examples can illustrate the approach: Galaxy Zoo or Tim Gower’s blog. However many cases of collaboration with the data-science community within competitions organised on Kaggle or Coda Lab platforms usually get limited by restrictions on those platforms. Common Machine Learning quality metrics do not necessarily correspond to the original research goal. Constraints imposed by the problem statement typically look artificial for ML-community. Preparing a perfect competition takes a considerable amount of efforts. On the contrary research process requires a lot of flexibility and ability to look at the problem from different angles. I’ll describe the alternative research collaboration process can bridge the gap between domain-specific research and data science community. Notably, it can involve academic researchers, younger practitioners and all enthusiasts who are willing to contribute. Such research process can be supported by an open computational platform that will be described along with essential examples for the audience of the workshop.
      Speaker: Andrey Ustyuzhanin
      Slides
    • 19:00
      School dinner
  • Thursday, 25 October
    • 16
      Machine learning outside HEP
      Speaker: Simone Frintrop
    • 10:30
      Coffee break
    • 17
      Machine Learning with Less or no Simulation Dependence
      Speaker: Benjamin Nachman
      Slides
    • 12:30
      Lunch break
    • 18
      Challenge
      Speakers: Gregor Kasieczka (Institut fuer Experimentalphysik / UHH), Dr Lydia Brenner (DESY)
      Slides
    • 15:30
      Coffee break
    • 19
      Challenge (continued)
  • Friday, 26 October
    • 20
      Advanced Deep Learning: Understanding the black box, physics, ...
      Speaker: Gregor Kasieczka (Institut fuer Experimentalphysik / UHH)
      Slides
    • 10:30
      Coffee break
    • 21
      Presentation of the best three challenge results
      notes
      Slides
    • 22
      Likelihood-free inference, its connections with a typical analysis and recent developments in ML
      Speaker: Gilles Louppe
      Slides
    • 23
      Goodbye
Powered by Indico v3.2.9