Deep neural networks have proven to effectively solve particular problems where a large amount of high quality training data is available. Supervised training of deep learning models utilizes mathematical operations to minimize or maximize an objective function, such as mean squared error or cross entropy. However, these models are unable to develop complex strategies. In contrast, by defining an environment, possible actions in this environment and by assigning rewards and penalties to these actions in different states reinforcement learning are capable of implicitly elaborating novel strategies.
Deep learning paradigm showed outstanding results in reinforcement learning. Recently a number of deep reinforcement learning strategies (like Deep Deterministic Policy Gradient Asynchronous Advantage Actor-Critic; Actor Critic using Kronecker-Factored Trust Region; etc. methods) evolved that achieve faster convergence than the baseline, Deep Q Network algorithm. Still, deep reinforcement learning faces all the difficulties of deep neural network design (data representation, hyperparameter optimization, etc.), and it requires a proper reward function design as well.
The goal of this PhD research is to elaborate novel algorithms towards elaborating complex and flexible strategies and understanding the possible current and future contexts and outcomes as well.
The effectiveness of the elaborated methods must be proven, at least in one application scenario. Such an application scenario can be (1) games, (2) intelligent chatbot, (3) model design, (4) robotics, etc.
The research can be conducted both in English and in Hungarian. For training the models public and private databases and high performance GPUs are available.
The possible research tasks of the PhD student are the following:
- Overview the related scientific papers, including the basic deep neural network elements, the basics and the recent results of reinforcement learning, focusing on deep reinforcement learning.
- Design and implement baseline methods of deep reinforcement learning in various application scenarios. Elaborate on a simulation environment.
- Conduct research on the potentials of deep reinforcement learning algorithms.
- Propose novel method(s) that elaborates complex and flexible strategies. Take into consideration the possible contexts and outcomes.
- Demonstrate the effectiveness of the results at least in one application scenario.
- Objective and subjective evaluation.
előírt nyelvtudás: english további elvárások: Programming and mathematical skills
felvehető hallgatók száma: 1
Jelentkezési határidő: 2020-06-15
2024. IV. 17. ODT ülés Az ODT következő ülésére 2024. június 14-én, pénteken 10.00 órakor kerül sor a Semmelweis Egyetem Szenátusi termében (Bp. Üllői út 26. I. emelet).
Bejelentkezés
