Marlos designs algorithms that learn abstractions for better credit assignment, generalization, and exploration in reinforcement learning.

Representation and reinforcement learning

Through his research, Marlos seeks to develop reinforcement learning methods that can be meaningfully used in real-world settings. He focuses on designing algorithms capable of learning abstractions that allow AI agents to tackle the three fundamental problems of reinforcement learning: generalization, exploration and credit-assignment. Currently, he is focused on designing theoretically-grounded algorithms that tackle these three problems concurrently. His research includes a number of different avenues, including designing representation learning methods tailored for reinforcement learning problems, developing AI agents that are capable of discovering temporally extended-behaviors (known as options), and creating systems capable of continual learning. Marlos is also passionate about reproducibility and proper experimentation in machine learning, having led several efforts on this topic in the past.

Marlos is a Canada CIFAR AI Chair at the Alberta Machine Intelligence Institute (Amii) and an assistant professor at the University of Alberta. Marlos's research interests lie broadly in machine learning, specifically in (deep) reinforcement learning, representation learning, continual learning, and real-world applications of all the above. He completed his B.Sc. and M.Sc. at UFMG, Brazil, and his Ph.D. at the University of Alberta. During his Ph.D., among other things, he popularized the idea of temporally-extended exploration through options, introducing the idea of eigenoptions. He was a researcher at DeepMind and at Google Brain for four years; during which time he made several contributions to reinforcement learning, including the application of deep reinforcement learning to control Loon's stratospheric balloons.

Marlos’ work has been published in the leading conferences and journals in AI, including Nature, JMLR, JAIR, NeurIPS, ICML, ICLR, and AAAI. His research has also been featured in popular media such as BBC, Bloomberg TV, The Verge, and Wired.

Marlos’s work has been published in the leading conferences and journals in machine learning, including Nature, JMLR, JAIR, NeurIPS, ICML and ICLR.