Sepp Hochreiter, head of the Institute for Machine Learning at the JKU, is globally renowned for his work on the vanishing gradient and his Long Short-Term Memory (LSTM) networks. These are considered important cornerstones of what has later become known as Deep Learning.
Research topics at the institute cover many fields of modern Deep Learning techniques including, among others, perception for self-driving cars, reinforcement learning, generative adverserial networks and toxicology prediction. Perception is an essential part of a self-driving vehicle and is being explored by developing new recurrent neural network architectures. Reinforcement learning models have vastly gained momentum over the last few years. At the institute, actor-critic systems are refined by applying LSTM-based reward distribution models. Generative adversarial network architectures are studied under the prospect of classical and quantum mechanical field theories. Many chemical compounds (drug candidates) fail in the late phases of the drug development pipeline. Therefore, at the institute Deep Learning and biclustering methods are being developed to identify new on-target and off-target effects in various drug design projects.
Potential Ph.D. Topics:
- Reinforcement Learning
- Drug discovery, target prediction, and toxicoloy
- Deep Learning for self-driving cars
- Deep Learning for generative models