Chris Williams: Research Interests

• Models for Understanding Time-Series/Condition Monitoring: Data that comes from a set of sensors recording through time can have rich structure. For example, patients in intensive care are monitored by many sensors. Our goal is to carry out condition monitoring, to identify different types of artifact and pathology in real time based on characteristic patterns in the data. See condition monitoring of premature babies for more details. This approach was extended to condition monitoring in an adult neuro ICU in the Southern General Hospital, Glasgow. Most recently we have studied how vital signs are affected by drug infusions.
• Image Interpretation. Object recognition can be cast in a statistical framework. This approach argues for image understanding using generative models, i.e. explaining an image by instantiated objects. My most recent vision work is in the "vision as inverse graphics" (VIG) framework, with PhD students Charlie Nash, Pol Moreno, and Lukasz Romaszko, see e.g. this paper, and this position paper.
  Work with Ali Eslami looked at the Shape Boltzmann Machine for modelling object shape and parts decompositions. Other work looked at lower-level edge-based and region-based models of images. The paper with Jyri Kivinen on Transformation Equivariant Boltzmann Machines considers learning to group edges while imposing rotation equivariance (i.e. the system will detect a specified configuration of edges at any rotation), and the paper with Nicolas Heess on Learning generative models texture models with extended Field-of-Experts learns higher-order random field models of visual texture. These contour- and region-based models can be combined to form models of textured regions.
  I was also one of the organizers of the influential PASCAL Visual Object Classes challenges concerning the recognition of object classes (e.g. cars, cats, etc) in images.
  In work with Michalis Titsias we learned sprite models of multiple objects that occur in a many images; for further information and movies etc click here. In earlier work (with Nick Adams, Steve Felderhof, Xiaojuan Feng, and Amos Storkey) we studied the use of tree-structured belief networks (TSBNs) and Dynamic Trees (DTs) as models of images. DTs are TSBNs that reconfigure themselves to a given input image or image sequence. Click here and here for futher information.
• Artificial Intelligence for Data Analytics: In the real world of data science, a very large fraction of the time taken on a project is taken up with data engineering or data wrangling. This includes data integration, data transformation, metadata discovery, structural variation, detection and repairing missing and anomalous data, and entity resolution. The goal of the AIDA project at the Alan Turing Institute is to draw on new advances in artificial intelligence and machine learning to produce technology that will help automate each stage of the data analytics process.
• Gaussian Processes. Since 1995 I worked on the use of Gaussian processes (GPs) for supervised learning along with several collaborators. An overview of this work can be obtained from the book Gaussian Processes for Machine Learning (C. E. Rasmussen and C. K. I. Williams, MIT Press, 2006). Subsequent work with Edwin Bonilla and Kian Ming Adam Chai focussed on multi-task learning for GPs, and work with Krzysztof Chalupka and Iain Murray addressed approximations for large-scale GPs.
• Unsupervised Learning: In addition to the work on learning learning multiple object from images, I also worked on GTM, the Generative Topographic Mapping (along with Chris Bishop and Markus Svensen). I have also worked on hierarchical mixture models, and probabilistic minor components analysis/extreme components analysis (with Felix Agakov and Max Welling). Recent work with Cian Eastwood is on the quantitative evaluation of disentangled representations.
• Applications:
  • PROTEUS, an EPSRC-funded IRC which aims to develop technology that will provide quick, in situ, in vivo diagnoses and management of lung diseases in the clinical environment.
  • Identifying Malaria Parasites in Microscopy Images. Work started with Carlos Sánchez Sánchez as a MSc project, in collaboration with John Quinn and his research group in Uganda.
  • Machine Learning for Compiler Optimization ( COLO, MILEPOST), with Mike O'Boyle and Edwin Bonilla.
  • Harmonising chorales in the style of J S Bach using HMMs (with Moray Allan). See also the HMM Bach demo.
  • Detecting satellite tracks in sky survey data (with Amos Storkey). We are also interested in other astronomical tasks, e.g. star/galaxy classification.
  • Haplotype reconstruction (with Michael Schouten).
  • Wind field modelling using Gaussian Processes, see Dan Cornford's pages on Neurosat at Aston Univeristy.
  • Modelling spike firing in oxytocin cells (with Duncan McGregor and Gareth Leng).

• Other: I am also interested in trying to understand the processing and representations used in animal visual systems.