CSP 2017 Online Program

Intro to Deep Learning with TensorFlow (303324)

Keywords: classification, deep learning, neural networks, Python, TensorFlow

Machine learning methods evolved in recent years with the speed of light. The conversation now gravitates toward deep learning methods and artificial intelligence. Deep learning methods shine when it comes to detecting weak signals in the midst of a lot of noise, a common occurrence in large datasets. Applications abound, from image processing to trading, video, speech, audio, text analytics and sentiment analytics. As the need for deep learning methods are ever increasing, while statisticians are often not exposed during degree programs to these methods but are now asked to use them, where should you start? This “Intro to Deep Learning with TensorFlow” session will give you a brief introduction to deep learning methods by focusing on a canonical example, the neural network with one hidden layer. Its classification performance will be compared to a logistic classifier on an image pattern recognition dataset. We will implement the models in TensorFlow and iPython Notebook. The TensorFlow toolkit is free to use via a Python environment and facilitates user applications of deep learning methods.