Deep Reinforcement Learning: Exploration with Neural Components in Diverse Operational Models

Zoom Research Seminar / 5th Floor EE Lecture 2

Past event — 14 February 2024

Spoken language

Profile image

Fatemeh Fakhredin

PhD Candidate

Kühne Logistics University - KLU

View profile


Machine Learning, along with its branches like Reinforcement Learning and Deep Reinforcement Learning, has gained significant attention lately for addressing various real-world challenges. Nevertheless, there exists some confusion regarding the appropriate application of different machine learning methods and the distinct advantages each offers. Fatemeh's research focuses on examining whether deep learning provides substantial benefits over conventional Reinforcement Learning for certain well-established problem categories such as inventory management problems. Deep Reinforcement Learning (DRL) stands out for employing deep neural networks to approximate components of reinforcement learning, such as the value function, policy, and model (state transition function and reward function). Essentially, Deep Reinforcement Learning merges artificial neural networks with a reinforcement learning framework, enabling software agents to effectively learn how to select their actions and achieve their objectives. In essence, it combines function approximation with target optimization, linking states and actions to the rewards they yield.


Fatemeh earned her Bachelor of Science in Industrial Engineering at Amirkabir University of Technology (AUT), specializing in Data Mining and Operations Research. She received "The Best Student Award" from the IE department for four consecutive years and graduated as the 2nd-ranked student among her peers in 2016. Admitted directly into the MSc program due to her exceptional talent, she secured a scholarship from Iran's National Elite Foundation for 2016 and 2017. In her Master dissertation, focusing on Systems Optimization, she developed a two-stage stochastic non-linear model for dynamic ride-sharing, addressing a vehicle routing problem. Her thesis, "Optimizing Dynamic Ride-Sharing to Reduce Traffic Jams in Large Cities," centered on implementing ride-sharing in Tehran. Graduating as the top student in 2018, Fatemeh later worked as a System Analyst and Developer at PSP Company in Iran, specializing in BPM-Software to enhance organizational efficiency and customer experience. She also expanded her expertise in data science and  machine learning with SQL Server and Python.