Enhancing Semiconductor Scheduling with Maskable PPO and Genetic Algorithms

June 17th 2025 from 11.45 – 13.15 in outdoor HS Kosuta (V.1.27 in case of rain)

Univ.-Ass. Dott. Mag. Peyman Eftekhari
Department of Artificial Intelligence and Cybersecurity (AICS)

Title
„Enhancing Semiconductor Scheduling with Maskable PPO and Genetic Algorithms“

Short Abstract:
Scheduling in semiconductor manufacturing presents significant challenges due to vast action spaces and dynamic production constraints. My doctoral research aims to develop advanced optimization strategies by integrating Maskable Proximal Policy Optimization (PPO) with Genetic Algorithms (GA). Maskable PPO enables efficient navigation of large discrete action spaces by allowing only valid actions, thereby accelerating the learning process and improving policy outcomes. Simultaneously, Genetic Algorithms are utilized to refine scheduling heuristics and establish strong baseline solutions. In this talk, I will discuss the rationale for combining reinforcement learning and evolutionary approaches, highlight the current progress, and share preliminary experimental findings alongside future research directions.

Short Bio:
I am a PhD candidate at the University of Klagenfurt under the supervision of Professor Martin Gebser. Alongside my research, I work as a university assistant and currently teach Algorithms and Data Structures in the bachelor’s program. My research focuses on the intersection of reinforcement learning and evolutionary computation, with a particular emphasis on optimization challenges in semiconductor manufacturing scheduling. I specialize in applying Maskable PPO and Genetic Algorithms to solve large-scale, complex scheduling problems. My academic background includes computer science and artificial intelligence, with practical experience in machine learning, optimization techniques, and teaching.