Raghavendra (https://www.linkedin.com/in/raghavendra-bommaraju-41aa995b) will talk about "Deep Learning & Neural Networks"
Artificial Neural Networks (ANNs) is a branch of Artificial Intelligence that has emerged with an influence from Biological Neural Networks and the goal to replicate them. Deep Learning is the process of learning an Artificial Neural Network. Deep Learning falls under the sub-branch of Artificial Intelligence called Representation Learning. Conventional ML algorithms are limited in their ability to process raw data and rather need domain specific features extracted from the data which can be hard to do.
Representation Learning works on raw data and solves this problem internally giving Deep Learning methods an edge over conventional ML methods. Although there has been work on ANNs since 1960s, they have gained momentum again in the recent few years. The reason of this being an increase of sizes of current datasets and a boost in computational power due to parallel processing in GPUs. Deep Learning has produced some record breaking results in the past few years in various problems including hand-written digit identification, and digital object recognition (ImageNet). In my talk, I would like to discuss about how Deep Learning achieves this success, why it is better than conventional ML algorithms for many problems, challenges involved in learning a deep neural network and optimized methods for deep learning, and newer modifications to conventional neural networks (CNNs, Auto-Encoders, etc.).