Thesis supervisor: János Abonyi
co-supervisor: Alex Kummer
Location of studies (in Hungarian): University of Pannonia Abbreviation of location of studies: PE
Description of the research topic:
In recent years, reinforcement learning (RL) has attracted significant attention from both industry and academia due to its success in solving some complex problems. The goal of the RL agent is to push the boundaries of what is currently possible through learning the optimal mapping of situations to actions (called policy) through a trial-and-error search guided by a scalar reward signal. In many challenging scenarios, actions affect not only the immediate reward, but also all subsequent rewards. These two features – guided trial-and-error search and delayed feedback – distinguish RL from all other topics of machine learning. If the investigated problem is high-dimensional or continuous states and/or actions are mandatory the traditional RL algorithms cannot work appropriately. In this case deep learning algorithms can help to identify the value function or the policies. The main application are of RL are in the logistics, robotics and self-driving cars. However, this technique can improve the efficiencies in chemical industry too, where it can be applied in many tasks, like decision making, scheduling, resource allocation or process control in any hierarchic level. The applicant’s task would be to discover the potentials of RL in chemical process control supported by different RL algorithms. For this purpose the applicant needs to perform process simulations which will serve as an environment for the RL agent.
Number of students who can be accepted: 1
Deadline for application: 2022-01-30
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).
Login
