Joonas Pussinen

Papers

Total Citations

H-Index

About

Joonas Pussinen is a researcher focused on artificial intelligence, particularly in the domains of behavioural cloning and reinforcement learning applied to video game environments. His most notable contribution, the 2020 paper "Benchmarking End-to-End Behavioural Cloning on Video Games," systematically evaluated how computers can learn to play games directly from human demonstrations, bypassing traditional reward-based training. This work provided a critical framework for comparing end-to-end approaches, where AI mimics human gameplay through raw pixel inputs, against more complex reinforcement learning methods. While his citation count is modest, the research addresses a foundational challenge in imitation learning—how to efficiently transfer human expertise to autonomous agents without extensive trial-and-error. Pussinen’s benchmarking methodology offers a standardized way to assess behavioural cloning performance, making his work valuable for students and researchers exploring accessible AI training techniques. His contributions help bridge the gap between human-like learning and machine efficiency, with potential applications extending beyond gaming to robotics and autonomous systems.