3:30 PM – 4:50 PM, Tuesdays and Thursdays, Filmor 355
Recent years have witnessed significant success on deep generative models and deep reinforcement, both are based on the deep-learning techniques in machine learning. Many methods have obtained state-of-the-art results on various real-world tasks, such as natural image generation and game playing.
In this seminar, we will review recent developments on two most proliferating topics: deep generative models and deep reinforcement learning. Topics includes variational autoencoders, generative adversarial neural networks, pixel RNN/CNN, deep Q learning, policy gradient methods and their variants.
The following topics will be covered:
1:00 PM – 3:00 PM, Mondays, Davis Hall 113A
Recent years have witnessed significant success of deep learning techniques in machine learning, obtaining state-of-the-art results on various real-world tasks, such as image classification, machine translation, image captioning and game playing with deep reinforcement learning.
The success of deep learning is partially due to the increase of computational ability, model complexity and algorithmic advantage. In this course, we will go over the three aspects in details. Specifically, we will cover a little bit of how to develop scalable deep neural models with excellent deep learning frameworks such as tensorflow; We will introduce basic concepts and techniques of deep learning; We will learn how to build deep learning models with feedforward neural networks, convolutional neural networks and recurrent neural networks, as well as some recent applications with different neural network architectures; Finally, we will cover algorithms for training deep neural network, including stochastic optimization and optionally stochastic gradient Markov chain Monte Carlo.
2:00 PM – 3:20 PM, Tuesdays and Thursdays, Norton Hall 214