HKHLR - HPC Hessen

HiPerCH 11 - Module 1

HiPerCH 11 - Module 1: Deep Learning

Module 1a: Introduction to Deep Learning

Topics

This course will introduce the concepts of Deep Learning. We will discuss the methods behind neural networks and how to use them set up non-linear models for solving e.g. classification tasks. The basic ideas involved in the process of training neural nets (backpropagation, minimization of loss functions by optimization methods) are treated. We shall further focus on the model evaluation and how to effectively train the model. The course will also cover advanced features of neural networks such as convolutional layers. In the hands-on session participants will train their own neural network using the Keras API of TensorFlow on a classical example of image recognition.

Method

  • Hands-on workshop

Target group and requirements

  • Beginners, basic Python knowlege will be of advantage.
  • Participants are expected to bring their own laptop with either Linux/MacOS or Windows with MobaXterm installed (see downloads for instructions). The hands-on session will be carried out through an SSH connection.
  • A WiFi connection will be provided via Eduroam (guest accounts are available).
  • This module is limited to 20 participants.

Module 1b: Scaling Deep Learning

Topics

In the second part of the course we will treat the basics of parallelizing computations with deep neural networks. Distributing the computational workload allows scaling of deep learning models. We will discuss different scaling approaches that are common when working with large deep learning models or big datasets. Participants will scale their own neural network (using the Keras API of Tensorflow and the distributed deep learning framework Hovorod) over multiple nodes on the Lichtenberg cluster. We will use an introductory example of speech recognition in order to demonstrate the scaling capability of a HPC cluster.

Method

  • Hands-on workshop

Target group and requirements

  • Users who participated in module 1a or have basic knowledge of Tensowflow/Keras.
  • Participants are expected to bring their own laptop with either Linux/MacOS or Windows with MobaXterm installed. The hands-on session will be carried out through an SSH connection.
  • A WiFi connection will be provided via Eduroam (guest accounts are available).
  • This module is limited to 20 participants.

 

Trainers

  • Marcel Giar (HKHLR)
  • Tim Jammer (HKHLR)

Date

  • Monday, September 23, 14:00-18:00 (module 1a)
  • Tuesday, September 24, 14:00-18:00 (module 1b)

Location

  • TU Darmstadt, Alexanderstraße 2, Karl-Plagge-Haus S1|22, Room 403 "New York"

Attendance fee

For each module part:

  • Students(Bachelor/Master): €2.50
  • PhD students and members of universities or public research institutes: €10.-
  • All other: €100.-

The fee includes coffee breaks and the evening event.


 

Participating Universities