Introduction to Deep Learning: A First Course in Machine Learning
Yosi Shibberu · 2018
Abstract Introduction to Deep Learning A First Course in Machine Learning Deep learning is a special type of machine learning that can take advantage of the growing availability of big data and the increasing computing power of GPUs. Deep learning is responsible for the dramatic improvements seen in recent years in speech and image recognition and is the key enabling technology behind self-driving cars. Companies like Google, Microsoft, Facebook and China's Biadu are investing heavily in research and development in deep learning. Applications of deep learning are growing rapidly. In this article, I describe an undergraduate course on deep learning that I am currently teaching for the first time at Rose-Hulman Institute of Technology. (I have taught 70% of the course.) Most of the students in the course are junior and senior computer science majors. Nearly all the students have not had a previous course in machine learning. The deep learning course is problem driven and builds on basic concepts students learn in calculus, statistics and probability. Key concepts from machine learning, e.g. the cardinal sin of overfitting, are introduced in the context of deep learning in a problem driven manner so that students discover and observe these concepts for themselves. Google's recently open-sourced deep learning software package, TensorFlow, is heavily used thoughout the course. Students are required to complete a deep learning project of their choice. A detailed concept map as well as online resources and data sets used for the course are described in the paper.