Gonçalo Guiomar
Research Fellow at the ETH AI Center, Zürich
Previously, Computational Neuroscience PhD at the Champalimaud Centre for the Unknown, Learning Lab
Researcher at Lunar Ring AI
Research
My research focuses on understanding the mechanisms through which organisms learn to construct representations of the world. With a background in physics, I employ methodologies and frameworks from neuroscience, mathematics, and machine learning to address these questions. I am presently a Research Fellow associated with the ETH AI Center in Zürich and the University of Zürich under the supervision of Valerio Mante and Benjamin Grewe. Previously I was a Ph.D. candidate in Joseph Paton's lab at the Champalimaud Centre for the Unknown in Lisbon. In parallel to my scientific research I have also developed a series of artistic collaborations that have allowed me to explore alternative means for articulating my research, with a particular emphasis in understanding autopoietic and semiotic processes through the construction of cybernetic systems both in the context of art installations and live performances.
Publications
"Representation learning over latent state policies" - Gonçalo Guiomar, Joseph J. Paton, Daniel McNamee (In Progress)
This project aims at understanding how acting in the world leads to a change in perception and how this could be adapted to a Reinforcement Learning framework based on Variational Inference.
"Action suppression reveals opponent parallel control via striatal circuits" - Bruno F. Cruz, Gonçalo Guiomar, Sofia Soares, Asma Motiwala, Christian K. Machens, Joseph J. Paton (2022). Nature, 607, 521-526
Built a multi-agent Reinforcement Learning model that is able to reproduce behavioral and neurophysiological data of mice in a time interval discrimination task and provide predictions for future experiments.
"Astrophysical Constraints on the Bumblebee Model" - Gonçalo Guiomar, Jorge Páramos (2014). Phys. Rev. D, D 90, 082002
Derived the strongest constraints for the existence of Dark Matter in the Sun using a symmetry breaking perturbation of Einstein's field equations.
Artistic Works
"Atavic Forest" within Latent Spaces (2022) and in collaboration with Jonathan Saldanha and Lunaring Art Colective (Tübigen).
This forest has been fully dreamt by Artificial Intelligence using state-of-the-art Generative Neural Networks trained with over 180,000 images in order to create a transversable space of visuals generated in real-time.
Teaching
Experience in tutoring students from all ages and education levels.
Master's thesis co-supervision @ Instituto Superior Técnico - "Diffusion Augmentation in Latent Program Spaces as a Cognitive Model of Psychedelic Action" - Carolina Caramelo (2023)
Introduction to Calculus @ INDP PhD Program (Spring 2022)
Reinforcement Learning @ INDP PhD Program (Spring 2019)
Introduction to Fourier Analysis @ CAJAL Computational Summer School (Summer 2018)
Introduction to Dynamical Systems @ INDP PhD Program (Spring 2017)
Programming: Python, C, C++, C#, Unity