Biography

Oliver Zeigermann is a Software Developer and Architect, Data Scientist, AI and Machine Learning Expert from Hamburg, Germany. He has developed software in many different languages and technologies over the past couple of decades, including C, C++, Java, Python, and JavaScript.

Workshop

Introduction to Deep Learning with TensorFlow

Deep Learning is a special and most promising variant of Supervised Machine Learning. Most recent break-throughs have been fueled by instead of programming a system, you instead use known data to train a system, like you do in deep learning. We will touch classic Neural Networks, Convolutional Neural Networks (CNNs) for image processing, and Recurrent Neural Networks (RNNs) for processing of texts and other sequences.

We will use TensorFlow with Keras-style Layers and provide notebooks hosted on Google’s Colab, that allow them to run on GPU. Thus there will be no need for any installation, all you need is a browser. We will use Python as our language, but you do not need any knowledge of it. Knowledge of any Object oriented language is sufficient.

Agenda

• Part I – Basics (Morning)
• Introduction to Machine Learning
• Introduction of our case study
• Generalization: The aim of Supervised Learning
• Different types of Machine Learning
• Understanding the Artificial Neuron and Neural Networks 
• How does an Artificial Neuron work?
• Activation functions
• Applying to our problem using TensorFlow, Keras on Colab
• Loss, Back Propagation and Optimizers
• Part II – Specialized Networks (Afternoon)
• Convolutional Neural Networks
• Applying Neural Networks to images
• How do convolutions work?
• Architectures for Convolutional Networks
• Case Study: Classifying Images using TensorFlow, Keras on Colab
• Recurrent Neural Networks
• Idea of RNNs and their advanced versions LSTMs and GRUs
• Embeddings for Natural Language Processing
• Case Study: Sentiment Analysis using TensorFlow, Keras on Colab

Course objectives

The main goal of this workshop is to introduce participants with main concepts of time series analysis, as well as with forecasting methods available in the PyFlux library.

Target audience

Level 1 – Introductory and overview material. Assumes little expertise with topic and covers topic concepts, functions, features, and benefits.

Course prerequisites

A personal computer. Attendees will need a laptop with a recent version of a Chrome or Firefox browser, no other installation necessary.