Skip to main content

Antonios Valkanas

PhD Student

renderFail

Hey! Here are a few things about me:

I was born in Athens, Greece and moved to Canada when I was 16. I finished my undergrad and started my Master’s in McGill University. Recently started my PhD as part of the ECE networks lab under the supervision of Prof. Mark Coates who also supervised my MSc.

In my undergrad I was supervised by Prof. Dennis Giannacopoulos of the COMPEM lab.

My main research interests include deep learning on graphs, semi-supervised learning, and Bayesian methods. Always interested in real world applications. In the past I have worked on computational electromagnetic problems and education in electrical engineering.

Personal Website (contains links to publications and project source code): https://antonvalk.github.io/

Current projects:

  • Negative sampling for incremental learning in recommendation systems
  • Neural architecture search via graph view of the design space using graph neural networks
  • Human motion prediction via Transformer for animation, see figure below for a rendering of our model

renderFail

Figure: Rendering of live motion from the test set inside the Unity engine. Comparison of human motion prediction. (Left): Our model. (Center): Linear interpolation. (Right): Ground Truth. Tracer lines represent time evolution of joint locations.