Special Seminar given by Prof. Markov of Aizu University

Deep Learning (DL) is an area of Machine Learning which has attracted much attention in recent years. It has significantly improved the performance of the computer technology in many fields such as signal processing, computer vision, natural language processing, robotics, econometrics, etc., and has been used to develop many "intelligent" applications. Bing Voice Search, Google Now, Skype Translator, and quite a few others are build using Deep Learning methods. It's been argued that DL is a big step towards achieving true Artificial Intelligence.

This talk gives a gentle introduction to Deep Learning starting with Machine Learning basics, Deep Learning principles and algorithms, application examples, and overview of the available software tools at the end.

Deep Learning is primarily associated with Deep Neural Networks (DNN), though deep or hierarchical structures based on other models exist as well. DNNs as a concept have been known for many years. However, just recently with the availability of large amounts of data, faster computers and the advances in learning algorithms their usage became feasible. The impressive performance DNNs deliver has been a surprise for many researchers. There are several types of DNNs such as Convolutional DNNs, Deep Recurrent Networks (DRNNs), Denoising Autoencoders (DAs), Long-Short Term Memory (LSTM) networks, which are going to be described in the talk. Their usage for various tasks will be explained with concrete examples.