Tutorial Abstract: High-level APIs such as Keras facilitate the development of deep learning models through a simple interface and enable users to train neural networks within a few lines of code. Building on top of TensorFlow, trained models can be exported and run efficiently on mobile devices. This enables a wide range of opportunities for researchers and developers. In this tutorial, we teach attendees three basic steps to run neural networks on a mobile phone: Firstly, we will teach how to develop neural network architectures and train them with Keras based on the TensorFlow backend. Secondly, we show the process to run the trained models on a mobile phone. In the final part, we demonstrate how to perform human activity recognition using existing mobile device sensor datasets.
Instructors: Huy Viet Le, Sven Mayer, Abdallah El Ali, Niels Henze
Date: 3rd September 2018
Location: MobileHCI ’18, Barcelona, Spain (https://mobilehci.acm.org)
Topics that we will cover:
- Introduction
- Overview of machine learning and recent advances in the field
- Hands-on introduction to Jupyter and Keras
- Discussion of performance metrics for classification and regression
- Bringing TensorFlow models to Android devices
- Human Activity Recognition research and supervised learning on time series data
- Hands-on training of human activity recognition classifier
- Deploying the classifier to Android devices
- Wrap-Up and Q&A