Summit 2014: Introduction to Deep Learning / Al Yaros

It's been a long time dream to make the machine a more human like, giving it an Artificial Intelligence. Teach it how to drive a car. Teach it how to identify objects in a given image for example give the computer an image captured in the front of a car and ask him if he can see pedestrians? Give the computer a web page and ask him what the page is talking about? Give the computer a set of metrics taken from patient blood readings and ask him to provide a diagnostic about the patient health? This is only a fraction of examples for machine learning applications there are many more fields such as speech recognition, recommendation systems, ranking and personalization of content all are often the basis of data products. There are many ways these days, to teach a machine – give it some data, ask it to learn the pattern and then apply this pattern over a new unseen data. Deep Learning is a new technique that outperformed many of the state-of-the-art algorithms in several research fields such as Audio, Text, and Vision. Artifacts from a decade of hand crafted human research turned out to be inferior to this new technique in these research fields. Deep Learning can be seen as The Rebirth of Neural Networks and as a Neural Network “fan” for many years I’m very excited to have the opportunity to see this rebirth. In this session I’ll be happy to share with you my passion to this field. I’ll give an introduction to machine learning via Neural Networks, I’ll explain what features are, what auto encoders are and what Deep Learning is all about and give you a taste from machine learning new frontier. MP3