Login

Deep Learning School "Basic Concepts"

Aug 8 – 12, 2022
Haus Nordhelle
Europe/Berlin timezone

contact

The Deep Learning School "Basic Concepts" is hosted by the community organization DIG-UM with support from the BMBF-funded ErUM-Data-Hub. The event serves the professional education of young scientists belonging to the ErUM Community.

In August 2022, we will host the first Deep Learning School at Landhaus Nordhelle in Meinerzhagen. In a comfortable ambience, intensive classes on Neural Network Building Blocks, Mastering Model Building, Convolutional Neural Networks, Recurrent Neural Networks and Introspection of Neural Networks will be held over 4 days. The workshop is aimed at deep-learning starters from all ErUM communities (Research on Universe and Matter) who have a basic knowledge of physics. 

A fee of 300€ will be charged for participation in the school. The workshop fee includes the cost of the workshop, accommodation and basic catering.

The workshop is full as the maximum number of 59 possible participants has been reached.

Please take note of the current Corona measures linked here.

 

 

Starts
Ends
Europe/Berlin
Haus Nordhelle
Meinerzhagen
Bulletin-1-20220805.pdf
Deep Learning School "Basic Concepts".pdf
ListOfParticipants-20220805.pdf

Topics

  • Neural Network Building Blocks (affine transformation, activation, loss function, mean-square-error, backpropagation, Universal Approximation Theorem, ...)
  • Mastering Model Building (input preprocessing, cross entropy, penalty term, stochastic gradient decent, learning strategies, batch normalization, residual learning,...)
  • Convolutional Neural Networks (convolutional layers, filters, parameter sharing, translational invariance, padding, pooling, striding, advanced concepts: inception, xception, unet, ...)
  • Recurrent Neural Networks (sequential relations, cell architectures, (bi-)directional analysis, long-short-term-memory, gated-recurrent-unit,...)
  • Introspection of Neural Networks (activation maximization, saliency maps, deconvolutional network, transposed convolutions, attribution-based predictions, discriminative localization, layer-wise-relevance-propagation...)

Lecturers

  • Dr. Caroline Heneka (University of Hamburg)
  • Jun.-Prof. Dr. Judith Reindl (UniBW Munich)
  • Dr. Nikolai Hartmann (LMU Munich)
  • Dennis Noll (RWTH Aachen University)
  • Dr. Marcel Rieger (University of Hamburg)

Tutors

  • Lars Sowa (Karlsruhe Institute of Technology)
  • Bogdan Wiederspan (University of Hamburg)
  • Boyang Yu (LMU Munich)

Prerequisite

  • Basic knowledge of physics, linear algebra

Fee

  • A fee of 300€ will be charged (includes the workshop, accommodation and catering). 

This workshop is supported by

Please take note of the current Corona measures linked here.

 

Add your participant profile for networking here: shorturl.at/aP457 (not mandatory)

Surveys
There is an open survey.
Powered by Indico v3.3.4