Time: July 4 (Wednesday) 16:20-17:20

Place: N902

Speaker: Alex Wozniakowski

Title: An Introduction to Supervised Deep Learning

Abstract: In this talk we discuss some elementary concepts in deep learning, which is a resurgent subfield of machine learning. We start by reviewing the basic recipe for constructing a machine learning algorithm, and we utilize this recipe to study the supervised learning problem. Deep learning algorithms are frequently designed to solve supervised learning problems, achieving state-of-the-art performance on benchmarks in computer vision, speech recognition, natural language processing (NLP), etc. We study the basic constituents of deep learning architectures, namely artificial neurons, such as the sigmoid family, softmax, and ReLU (rectified linear unit). Moreover, we study agglomerations of artificial neurons into deep artificial neural networks, which have neuroscientific and biological inspiration.