Deep learning is a machine learning technique that uses vast volumes of data and complex algorithms to teach computers to do what comes naturally to humans. As Canada Research Chair in Machine Learning, Dr. Graham Taylor’s aim is to achieve algorithmic and application advances in deep learning.
He and his research team have three main objectives. First, they aim to extend the benefits of generative models beyond digital assets like images and text. They are also developing graph generative models that automatically discover and assemble parts. Secondly, they are developing a foundation model for DNA barcode data to support biodiversity science. Foundation models are very large models trained on massive datasets that can easily be adapted to downstream tasks. Lastly, Taylor and his team will reduce the carbon footprint of deep nets—networks that process data in complex ways—while improving their generative and predictive abilities.