Workshop / Overview

Note:
Workshop documents are available on Github
(full notebooks and slides).

This workshop gives a quick overview over Tensorflow, from basic concepts and low-level operations all the way through using predefined estimators and running the model in the cloud. The main part of the workshop will be spent on reading through carefully prepared example code and asking questions. This allows to cover a lot of material, but post-workshop efforts are certainly needed in order to master the content.

ML knowledge is not strictly required for understanding the Tensorflow code, but we won't have time to explain ML theory during
the workshop.

Workshop / Outcome

Participants will
- Data preparation for Tensorflow and create a sharded dataset
- Understand Tensorflow basics (graph, execution model and shapes)
- Connect data to estimator, use LinearClassifier, DNNClassifier
- Train and deploy model in the cloud

The workshop material covers many more topics (convolutions, Keras, recurrent networks) and workshop participants will be well-prepared to tackle these advanced topics after the workshop.

Workshop / Difficulty

Beginner level

Workshop / Prerequisites

  • Python knowledge
  • Laptop with pre-installed Docker CE (some Windows users might need to use Docker Toolbox instead)
  • After installing Docker execute the following command to fetch the workshop image: "docker pull andstein/tensorflow-basics"
  • Make sure that you can run and access the notebook contents in the browser
  • For users who fail to install Docker (sad!) we will provide a virtual machine during the workshop

Track / Co-organizers

Bartek Wołowiec

Software Engineer, Google

Ruslan Habalov

Security Engineer, Google

Andreas Steiner

Software Engineer, Google

AMLD EPFL 2018 / Workshops

TensorFlow Basics 2018 – Saturday

With Bartek Wołowiec, Ruslan Habalov & Andreas Steiner

09:00-12:00 January 274ABC

Financial Predictions with Machine Learning

With Stefano Tempesta

13:30-16:30 January 275BC

Open Food Hackdays 2018

With OpenData.ch, Hannes Gassert & Nikki Böhler

09:00-17:00 January 28

AMLD / Global partners