Deep Learning- Basic Concepts

 

Workshop Title

Deep Learning- Basic Concepts

 

Workshop Abstract

Data representation is a crucial task in machine learning applications so that their success depends on the representation quality. Recently, Deep learning techniques could achieve the best rank in most machine learning applications due to their power in providing robust and discriminative representations for the data at hand. These methods could significantly improve the power of machine learning techniques in solving the problems which were poorly handled by existing shallow methods such as speech recognition, image colorization, image captioning, etc. The Deep Learning phenomenon is actually nothing more than the well-known neural networks with several number of layers which makes them able to provide several levels of abstractions and improve the generalization ability. There was a drawback in basic Neural networks before 2003 because of the problem in their training process called “vanishing gradient problem” which was solved by Hinton in 2003 and the Deep learning revolution appeared in machine learning field so that  large number of last publications are accounted for Deep learning field.

 

Workshop Outlines

  1. Motivation
  2. Representation Learning
  3. Basic NN Concepts
  4. RBM & relU
  5. Basic Deep Structures

 

Workshop Organizers*

Name

Affiliation

Contact

Corresponding

Dr Zohreh Azimifar

Associate Professir- Computer Engineering Dpt, Shiraz university

 

 

Partisipants in this workshop must have these requirements on their laptops in order to try out workshop instructed materials in class.

Anaconda 4.4.0
Keras 2.0.6 
Tensorflow 1.2.1
Python 3.6.1