About me

I am a Research Assistant at the University of Alberta working with Marlos C. Machado and Michael Bowling. I am interested in understading the computational principles that explain the emergence of intelligent systems. In particular, I am focused in the problems of continual exploration and state construction in reinforcement learning. These two are inseparable problems whose solution encompass many other sub-problems and related areas like option and sub-goal discovery, representation learning and spectral-based algorithms, partial observability and memory, hippocampal neural activation, distance and graph learning, planning, continual learning, and online learning.

In my recent work, I concentrate on analyzing and comparing spectral based representations with the average activation of cells in the hippocampal formation (see Stachenfeld et al., 2017 for a reference). Also, I am working in extensions of the Laplacian representation in asymmetric environments.


Publications

Second Laplacian eigenvector in multiple grid environments.

Proper Laplacian Representation Learning
Diego Gomez, Michael Bowling, Marlos C. Machado
International Conference on Learning Representations, 2024
pdf | code

Second Laplacian eigenvector in multiple grid environments.

Information Optimization and Transferable State Abstractions in Deep Reinforcement Learning
Diego Gomez, Nicanor Quijano, Luis Felipe Giraldo
IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022
pdf | code