Abstract:
Deep generative models are powerful tools for modeling data blending the representational power of neural networks with the flexibility of Bayesian networks. First, I'll describe PolyODE, a Neural ODE that models the latent continuous-time process as a projection onto a basis of orthogonal polynomials. This formulation enforces long-range memory and preserves a global representation of the underlying dynamical system. Our construction is backed by favourable theoretical guarantees and in a series of experiments, we demonstrate that it outperforms previous works in the reconstruction of past and future data, and in downstream prediction tasks. Unsupervised learning is often used to uncover clusters in data. However, different kinds of noise may impede the discovery of useful patterns from real-world data. Next, I’ll describe SubLign , a deep generative, continuous-time model of time-series data that clusters time-series while correcting for censorship time. I’ll highlight under which clusters and the amount of delayed entry may be identified from data under a noiseless model. On real-world clinical datasets of heart failure and Parkinson’s disease patients, I’ll showcase how interval censoring can adversely affect the task of disease phenotyping and how SubLign corrects for these sources of error and recovers known clinical subtypes.
Rahul G. Krishnan is an Assistant Professor of Computer Science and Medicine (Laboratory Medicine and Pathobiology). He is a Canada CIFAR AI Chair at the Vector Institute and a member of the Temerty Center for Artificial Intelligence in Medicine. His research develops algorithms to tackle problems in healthcare such as clinical decision making and deployable, trustworthy machine learning. He was previously a Senior Researcher at Microsoft Research New England. He received his MS from New York University and his PhD in Electrical Engineering and Computer Science from MIT.
