Artificial intelligence scientist
I was featured for the second time on Machine Learning Street Talk. This interview was shot at NeurIPS 2023 last year, where I was presenting our work on generalized implicit neural representations from my time at EPFL. Cheers!
I had the pleasure of being a guest on Machine Learning Street Talk to chat about cellular automata, emergence, life, the universe, and my own work on graph neural cellular automata. I had a great time with Tim and Keith, they are doing an incredible work with the podcast and it’s really an honor to having been a part of it. Enjoy! P.S. I was so nervous and hyper-excited that I lost my own train…
Cellular automata (or CA for short) are a fascinating computational model. They consist of a lattice of stateful cells and a transition rule that updates the state of each cell as a function of its neighbourhood configuration. By applying this local rule synchronously over time, we see interesting dynamics emerge. For example, here is the transition table of Rule 110 in a 1-dimensional binary CA:…
This is Part 2 of an introductory lecture on graph neural networks that I gave for the “Graph Deep Learning” course at the University of Lugano. After a practical introduction to GNNs in Part 1, here I show how we can formulate GNNs in a much more flexible way using the idea of message passing. First, I introduce message passing. Then, I show how to implement message-passing networks in…
This is Part 1 of an introductory lecture on graph neural networks that I gave for the “Graph Deep Learning” course at the University of Lugano. At this point in the course, the students had already seen a high-level overview of GNNs and some of their applications. My goal was to give them a practical understanding of GNNs. Here I show that, starting from traditional CNNs and changing a few…
Yesterday, it was board game day at the lab where I have been working recently. Everyone got together for lunch at Snakes & Lattes, a Torontonian board game cafè chain, and we spent a couple of hours laughing and chatting and, obviously, playing board games. The lab has a go-to traditional game for the occasion: Telestrations. The game is inspired by the classic childhood’s game of Chinese…
In this post, I’m going to summarize some conceptual problems that I have found when comparing different graph neural networks (GNNs) between them. I’m going to argue that it is extremely difficult to make an objectively fair comparison between structurally different models and that the experimental comparisons found in the literature are not always sound. I will try to suggest reasonable…
Mathematically modeling how epilepsy acts on the brain is one of the major topics of research in neuroscience. Recently I came across this paper by Oscar Benjamin et al., which I thought that it would be cool to implement and experiment with. The idea behind the paper is simple enough. First, they formulate a mathematical model of how a seizure might happen in a single region of the brain. Then,…
In our latest paper, we presented a new pooling method for GNNs, called MinCutPool, which has a lot of desirable properties as far as pooling goes: It’s based on well-understood theoretical techniques for node clustering; It’s fully differentiable and learnable with gradient descent; It depends directly on the task-specific loss on which the GNN is being trained, but … It can be trained on its own…
The first paper on which I worked during my PhD is about detecting changes in sequences of graphs using non-Euclidean geometry and adversarial autoencoders. As a real-world application of the method presented in the paper, we showed that we could detect epileptic seizures in the brain, by monitoring a stream of functional connectivity brain networks. In general, the methodology presented in the…