AI, Data, and Machine Learning

TH10 Introduction to the CNTK Neural Network Library


11:00am - 12:15pm

Level: Introductory to Intermediate

James McCaffrey

Research Software Engineer, Microsoft Research


Microsoft CNTK is an extremely powerful code library that can be used to create deep neural network prediction models. In this informal, but in-depth presentation, Dr. James McCaffrey from Microsoft Research will explain exactly what CNTK is, and describe the types of problems CNTK can, and cannot solve. You will leave this presentation with all the information you need to create a working deep neural prediction system. Topics covered include: installing CNTK, the CTF data file format and data readers, creating deep neural network layers, understanding squared error vs. log loss, training with advanced stochastic gradient descent, model evaluation, and transferring a CNTK model to a C# program.

You will learn:

  • Exactly what CNTK can and cannot do
  • How to create a deep neural prediction system using CNTK
  • How to transfer a CNTK model to a C# program