Building neural networks with Linux

This course is perfect for those wanting to make practical use of the world of possibilities opened by modern Artificial Intelligence evolutions.

The course starts by looking at the various Neural Network implementations and provides the minimal theoretical background needed to understand it. We look at implementation choices. We learn to use the most common implementations with Python and Keras. We use the Neural Compute stick with build in hardware accelerated Tensorflow network to implement in combination with a Raspberry Pi and a camera practical object recognition. We explore pre-build neural networks and learn how to build them. And apply our knowledge in a Data Science fashion too. Finally we look at limitations, debugging and topics of interest.

During this course all participants will have the opportunity to build and experiment with a multifunctional small-footprint embedded target with powerful hardware optimised neural network capabilities. After the course, the participants can take the board with them to continue experimenting.

  • understand the various Neural Network implementations
  • learn to use the most common implementations with Python and Keras
  • apply this to a practical visual object recognition task
  • learn how to build neural networks
  • apply this knowledge in a Data Science fashion

Schedule a training?

Delivered as a live, interactive training: available in-person or online, or in a hybrid format.

REQUEST IN-COMPANY TRAINING

 

Public training calendar

No public sessions are currently scheduled. We will be pleased to set up an on-site course or to schedule an extra public session (in case of a sufficient number of candidates). Interested? Please let us know.

Intended for

Everybody who is responsible for designing and building Artificial Intelligence Systems.

Background

IT Background and general Linux skills and command line experience (see Linux / UNIX fundamentals). Python knowledge is not required, but can be helpful during the course (see Python fundamentals). Programming experience is however essential to better understand the concepts.

You can test for yourself to see if you have enough background by filling out the on-line self-test "Linux" and the online selftest "Introduction to programming".

Main topics

  • Introduction
    • An in-depth look at Neural Networks and Python
    • An introduction to Neural Networks and neural network types
    • Convolutional Neural Networks
    • Python fundamentals
    • Set up of our development board: Raspberry Pi, Camera and Neural Compute Stick
  • Object Recognition
    • Implementation of object recognition with the Neural Compute Stick and Tensorflow
  • Sequential model and API, Compilation, training and layers with Keras
    • Multilayer Perceptron (MPL)
    • Convolutional Neural Networks (CNN)
    • Recurrent Neural Networks (RNN)
  • Using pre-trained neural networks
    • Background
    • A practical look at using existing pre-trained neural networks
    • Implementation of a self-driving car platform
  • Neural Networks for Data Science
    • Background
    • Example: Value Estimations
  • Limitations, debugging and Topics of interest
    • We make time to look at the practical challenges of the course participants.

Training method

Course/workshop: classical education with practical exercises.

Course materials provided, complemented with 1 book: Tutorial: Building Neural Networks with Linux, by Jasper Nuyens a free ARM-based Embedded Linux board with camera and Intel NCSM2450.DK1 Movidius Neural Compute Stick and a small self-driving car platform.

Certificate

At the end of the session, the participant receives a "Certificate of Completion".

Duration

5 days.

Course leader

Linux Belgium.


SESSION INFO AND ENROLMENT