Introduction to Deep Learning

Original article was published on Deep Learning on Medium

Introduction to Deep Learning

For the past few years, the past decade or so, deep learning has become very ubiquitous and it has found applications in a wide range of areas such as speech, computer vision, natural language processing. Most of the state of the art systems in these areas from even companies like Google, Tesla, Facebook, etc., use Deep Learning as the underlying solution.

Deep Learning is also enabling brand new products and businesses and ways of helping people recreate it. Everything ranging from better healthcare where deep learning is helpful in reading X-Ray Images to delivering personalized education to better agriculture and even self-driving cars and many more.

For understanding Deep Learning in a naive approach, we will first define Intelligence: as the ability to process information to inform future decisions.

Artificial Intelligence is simply the field that focuses on building algorithms that can process information to inform future decisions.

AI is the new Electricity as we may put it in simpler words. Electricity had transformed may industries like transportation, manufacturing, healthcare, etc. AI will now bring about an equally big transformation.

Machine Learning: It is a subset of Artificial Intelligence that takes this idea a step forward and specifically teaches an algorithm to do this task at hand without being explicitly programmed.

Deep Learning: It is a subset of Machine Learning which takes this idea even a step more further and helps to extract useful pieces of information needed to perform future predictions or make a decision.

© Alexander Amini and Ava Soleimany
MIT 6.S191: Introduction to Deep Learning
IntroToDeepLearning.com

Over the last couple of decades with the digitization of society, we have collected more and more data than ever before. Every activity that we indulge on the Internet today marks a digital footprint that is going to stay in cyberspace forever.

Looking at Traditional Machine Learning Algorithms ( SVM, Decision Trees, Logistic Regression), the performance of most of them would plateau, even as we feed more and more data into it. But what we found out recently is that if we train a small neural network, the performance of the system increases and it keeps on increasing up to a certain limit if we start training larger size Neural Networks. GPU computing helped us in effectuating this holy grail by speeding up the process of training larger neural networks.

Stanford CS230: Deep Learning