SIAM@Purdue is holding an introductory workshop in the area of machine learning this summer. The workshop is open to all current students and faculty and will focus on neural networks and their implementation in TensorFlow.
Video recordings for each of the workshop talks are available on YouTube.
Examples of complete TensorFlow models are also available on GitHub.
Abstract: The TensorFlow software library provides a state-of-the-art framework for high-performance, scalable machine learning. While this library allows developers and researchers to create and customize highly complex models, getting started with TensorFlow can be quite overwhelming. In this workshop we will walk through the key steps for building and training models with TensorFlow, highlighting the best practices laid out in the TensorFlow performance guide.
Abstract: Artificial neural networks have established themselves at the core of modern machine learning research. In this workshop we will investigate the key components of neural network architectures and analyze the optimization algorithms used to train them. In addition, we will examine the computational costs of common network layers and see how collaborative filters can be designed to construct fast, high-performance networks.
Abstract: The field of deep learning has grown immensely in recent years, emitting a constant flow of novel architectures and training strategies. In this workshop we will survey three fundamental classes of deep learning models. To begin, we will take a brief look into designing classifier networks; we will then proceed to cover the theory and TensorFlow implementation details for two network architectures at the foundation of modern deep learning research: Generative Adversarial Networks (GANs) and Variational Autoencoders (VAEs).