Ilya Loshchilov

École Polytechnique Fédérale de Lausanne, Université Paris-Sud

Papers

Total Citations

H-Index

About

Ilya Loshchilov is a leading researcher in machine learning, evolutionary computation, and optimization, best known for his groundbreaking contributions to neural architecture search and automated machine learning. His most-cited work, "RoboGen: Robot Generation through Artificial Evolution" (2014), with 72 citations, pioneered the use of evolutionary algorithms to autonomously design and optimize robotic systems, demonstrating how artificial evolution can generate efficient robot morphologies and controllers. This work has significantly advanced the field of evolutionary robotics, offering a scalable framework for automating robot design. Loshchilov’s research also includes "Fast Adaptive Object Detection towards a Smart Environment by a Mobile Robot" (2013), which, despite fewer citations, showcases his early work in real-time perception systems. His broader impact is evident in his development of the AdamW optimizer and decoupled weight decay, which have become standard tools in deep learning, widely adopted for training large-scale neural networks. With thousands of citations across his portfolio, Loshchilov’s work bridges evolutionary algorithms and deep learning, influencing both theoretical optimization and practical AI applications. His achievements underscore a career dedicated to automating intelligence through evolution and learning.