• Publications
  • Influence
Ranked Reward: Enabling Self-Play Reinforcement Learning for Combinatorial Optimization
TLDR
Results from applying the R2 algorithm to instances of a two-dimensional and three-dimensional bin packing problems show that it outperforms generic Monte Carlo tree search, heuristic algorithms and integer programming solvers.
Early Computational Detection of Potential High Risk SARS-CoV-2 Variants
TLDR
A COVID-19 Early Warning System combining structural modelling with AI to detect and monitor high risk SARS-CoV-2 variants, identifying >90% of WHO designated variants on average two months in advance.
More unlabelled data or label more data? A study on semi-supervised laparoscopic image segmentation
TLDR
A laparoscopic liver image segmentation application is investigated by altering the quantities of labelled and unlabelled training data, using a semi-supervised segmentation algorithm based on the mean teacher learning paradigm, which reports a significantly higher segmentation accuracy, compared with supervised learning.
DeepReg: a deep learning toolkit for medical image registration
DeepReg (this https URL) is a community-supported open-source toolkit for research and education in medical image registration using deep learning.
Learning image quality assessment by reinforcing task amenable data selection
TLDR
This work shows that the controller-predicted image quality can be significantly different from the task-specific image quality labels that are manually defined by humans, and demonstrates that it is possible to learn effective image quality assessment without using a “clean” validation set.
Longitudinal Image Registration with Temporal-order and Subject-specificity Discrimination
TLDR
A learning-based image registration algorithm to quantify changes on regions of interest between a pair of images from the same patient, acquired at two different time points is described and a novel regularisation method based on maximum mean discrepancy, between differently-sampled training image pairs is proposed.
Adaptable image quality assessment using meta-reinforcement learning of task amenability
TLDR
This work demonstrates that the IQA agents pre-trained on non-expert task labels can be adapted to predict task amenability as defined by expert task labels, using only a small set of expert labels.
Ranked Reward: Enabling Self-Play Reinforcement Learning for Bin packing
TLDR
The Ranked Reward (R2) algorithm is presented which accomplishes this by ranking the rewards obtained by a single agent over multiple games to create a relative performance metric and uses a reward ranking mechanism to build a single-player training curriculum that provides advantages comparable to those produced by self-play in competitive multi-agent environments.
FEW-SHOT Image Segmentation for Cross-Institution Male Pelvic Organs Using Registration-Assisted Prototypical Learning
TLDR
This work presents the first 3D few-shot interclass segmentation network for medical images, using a labelled multi-institution dataset from prostate cancer patients with eight regions of interest, and proposes an image alignment module registering the predicted segmentation of both query and support data to a reference atlas space.
...
...