Stephanie Hyland, MASt. Cambridge University (Applied Mathematics and Theoretical Physics)

"Explore the world. Nearly everything is really interesting if you go into it deeply enough." - Richard Feynman

PhD Student

E-Mail
hyland@get-your-addresses-elsewhere.inf.ethz.ch
Phone
+41 44 632 23 71
Address
ETH Zürich
Department of Computer Science
Biomedical Informatics Group Universitätsstrasse 6
CAB F52.1
8092 Zürich
Room
CAB F53.1

My research focuses on the application and development of machine learning in healthcare.

I am interested in time series models (such as recurrent neural networks and Gaussian processes) appropriate for modelling physiological signals, and phenotyping through representation learning. I am also interested in the use of reinforcement learning techniques in healthcare.

I studied theoretical physics in Trinity College Dublin for my undergraduate degree, where I focused on lattice field theory. In 2012 went to Cambridge University (St. John’s College) to do Part III of the mathematical tripos in applied mathematics and theoretical physics. I then moved to New York to join the Tri-Institutional Training Program in Computational Biology. I spent a year at Cornell University before joining the Rätsch lab at Memorial Sloan Kettering Cancer Center in New York City, and relocated to Switzerland in 2016 when the group moved to ETHZ.

Abstract Generative Adversarial Networks (GANs) have shown remarkable success as a framework for training models to produce realistic-looking data. In this work, we propose a Recurrent GAN (RGAN) and Recurrent Conditional GAN (RCGAN) to produce realistic real-valued multi-dimensional time series, with an emphasis on their application to medical data. RGANs make use of recurrent neural networks in the generator and the discriminator. In the case of RCGANs, both of these RNNs are conditioned on auxiliary information. We demonstrate our models in a set of toy datasets, where we show visually and quantitatively (using sample likelihood and maximum mean discrepancy) that they can successfully generate realistic time-series. We also describe novel evaluation methods for GANs, where we generate a synthetic labelled training dataset, and evaluate on a real test set the performance of a model trained on the synthetic data, and vice-versa. We illustrate with these metrics that RCGANs can generate time-series data useful for supervised training, with only minor degradation in performance on real test data. This is demonstrated on digit classification from 'serialised' MNIST and by training an early warning system on a medical dataset of 17,000 patients from an intensive care unit. We further discuss and analyse the privacy concerns that may arise when using RCGANs to generate realistic synthetic medical time series data.

Authors Stephanie L Hyland, Cristobal Esteban, Gunnar Rätsch

Submitted arXiv

Link

Authors Paulina Grnarova, Florian Schmidt, Stephanie L Hyland, Carsten Eickhoff

Submitted NIPS Workshop on Machine Learning for Healthcare

Link

Authors Stephanie L Hyland, Theofanis Karaletsos, Gunnar Rätsch

Submitted NIPS Workshop on Machine Learning for Healthcare, 2015

Link

Authors Charles G Danko, Stephanie L Hyland, Leighton J Core, Andre L Martins, Colin T Waters, Hyung Won Lee, Vivian G Cheung, W Lee Kraus, John T Lis, Adam Siepel

Submitted Nature Methods