The path to Deep Learning with CNTK (ENG)

Pablo Doval

Date: November 29, 2017

Location: University of Applied Social Sciences. Kalvarijų str. 137E, LT-08221 Vilnius, Lithuania

About the workshop

Abstract
Learn the basics of Deep Learning models for computer vision in a hands-on way

Agenda

• Why do we need to create our own models?
• Introduction to Deep Learning
• The “Hello World” of CNTK: Logistic Regression
• Logistic regression applied to Computer Vision
• Multi-Layer Perceptron approach
• Convolutional Neural Networks: At last! Real Deep Learning
• Other topologies: autoencodes, LSTM, etc.
• Other computer vision models: ResNet, FastRCNN
• Beyond Computer Vision

Target Audience
Developers interested in building deep learning models, and researchers interested in comparing the specific implementation with other frameworks.
No previous experience is required, as all concepts will be introduced in the theory modules of he workshop; however, a minimum knowledge of ML concepts and practices (such as understanding the train/test/validation cycle, etc…) would be beneficial.

Practice
The following labs will be done during the course of the workshop:
• Environment set up
• Basic Logistic Regression
• MNIST classifier (guided): Logistic Regression, MLP, CNN
• Playing with the hyperparameters: Minibatch sizes, Learning Rates
• MNIST classifier challenge (your turn!)

Requirements
We will perform the installation of the required wheels for using CNTK as part of the labs, but having the following pre-requisites installed will save time and potential issues during the workshop:
• Anaconda distribution with Python 3.5 environment
• Python IDE (VSCode recommended)
• Git client

Trainer
Pablo Doval is the Data Team Lead at Plain Concepts, where he has helped architect and implement some really exciting big data projects during the last years, having specialized in Microsoft’s Data Platform and ML/AI. Previously to that, he worked at Microsoft’s GTSC, in SQL Server’s Team.
This workshop will not just be an opportunity to learn hands-on the basic concepts of Deep Learning and the specific implementation on CNTK, but will also provide the attendants with the opportunity to learn from Pablo’s and Plain Concept’s experience in the field and real life projects.