Home
Home
Contact
My News
Research
Publications
Teaching
Students
Service
CV
A mugshot at the Hundertwasserhaus in Vienna

Joshua Knowles

Visiting Scholar, Machine Learning and Optimization Group


Contact | Office hours

 

Note: (June 1, 2015) I have now moved to the University of Birmingham. I will have a webpage there shortly.

About

The chief research interests of my group are evolutionary computation, artificial life and related areas in computer science, mathematical modelling and biology.

Other key topics are expensive global optimization, automation of experimental science, and links between machine learning and optimization. The themes of multiobjective optimization and decision analysis also feature in much of our work. More

Some rather more focused (and technical) research highlights, for those interested, can be seen below.

Earlier research highlights in the words of others

Hypervolume maximization (2002): "Knowles and Corne were the first to propose the integration of the hypervolume indicator into the optimization process. In particular, they described a strategy to maintain a separate bounded archive of nondominated solutions based on the hypervolume indicator", J. Bader (2009)

Multiobjective memetic algorithms (2000): "Knowles and Corne proposed a greedy local search method mainly based on dominance relation. Their idea is to accept a new neighborhood solution if it dominates the current solution... An obvious advantage of dominance relation is its independence on any monotonic transformation of objective functions." A. Jaszkiewicz et al. (2012)

Adaptive grid-based archiving (1999): "Since the procedure is adaptive, no extra parameters are required [...] . This adaptive grid (or variations of it) has been adopted by several modern multiobjective evolutionary algorithms." C. A. C. Coello (2006).

PAES algorithm (1999): "... the most competitive algorithm" [concerning scalability in decision space, and] "...the second best algorithm in terms of speed." J. J. Durillo et al. (2008), an empirical comparison of scalability of MOEAs.

Multiobjectivization (2001): "...researchers proposed the so-called 'multiobjectivization' by which a single-objective optimization problem is decomposed into several subcomponents considering a multi-objective approach (Jensen, 2003), (Knowles et al., 2001). This procedure has been found to be helpful in removing local optima from a problem and has attracted a lot of attention in the last few years." C. A. C. Coello (2006)

ParEGO (2006): "In ParEGO, the nondifferentiabilities are smoothed out by the surrogate model, making the actual EI criterion continuous and differentiable. ... For ParEGO, only one model has to be computed making it the fastest of all approaches." T. Wagner et al. (2010)
"Recently, some researchers have proposed the use of black-box optimization techniques normally adopted in engineering to perform an incredibly low number of fitness function evaluations while still producing reasonably good solutions (see for example (Knowles, 2006))." C. A. C. Coello (2006)

 

Research key terms: evolutionary computation, heuristics, combinatorial optimization, multi-objective optimization, swarm intelligence, sequential design of experiments, active learning, reinforcement learning, surrogate modeling, cross-disciplinary work in the bio-sciences.

What's happening now? ...see My News.

Qualifications

BSc (Hons); PGCE; MSc (Dist'n); PhD; FHEA

Awards and Fellowships

  • Outstanding IEEE Transactions on Evolutionary Computation Paper Award (2006)
  • BBSRC David Phillips Fellowship (2002-2008)
  • FNRS Chargé de Recherche (Fellowship of the Belgian National Science Fund) (declined 2003)
  • Outstanding IEEE Transactions on Evolutionary Computation Paper Award (2003)
  • EC Marie Curie Posdoctoral Fellowship (2001-2003)

Industrial Collaboration

My PhD was sponsored by BT. Other projects were carried out in collaboration with HBOS, Astra Zeneca, GSK, Waters, Thermo Instruments, and Combimatrix. I did a sabbatical at Theo Chocolate, Seattle and The University of Washington in 2009.

Grant Income

Latest funding:
EPSRC internal: Constrained global optimization for fragment-assembly approaches to protein structure prediction (co-I)
BB/I023755/1 MUSCLE: Multi-platform Unbiased optimization of Spectrometry via Closed-Loop Experimentation (PI)
BB/C008219/1 MCISB: The Manchester Centre for Integrative Systems Biology (co-I)

Previous funding:
BBS/A/00013 BBSRC David Phillips Fellowship: Interactive evolutionary search for post-genomic knowledge discovery and prediction using GRID computing (PI)
BB/C007158/1 Constrained optimisation of metabolic and signalling pathway models: towards an understanding of the language of cells (co-I)
BB/C519038/1 HUSERMET: The human serum metabolome in health and disease (co-I)
EP/D013615/1 A convergent strategy for high efficiency quantitative proteomics (co-I)

People

We have (had) collaborations with the following people (with apologies for any omissions).
  • David Corne (Heriot-Watt)
  • Julia Handl (Manchester Business School)
  • Douglas Kell (Manchester Institute for Biotechnology)
  • Richard Allmendinger (UCL)
  • Steve O'Hagan (MIB)
  • Ben Stappers (Jodrell Bank)
  • John Brooke (IMG / Research Computing Services)
  • Richard A. Watson (Southampton)
  • Carlos Fonseca (U. Coimbra)
  • Eckart Zitzler (PH Bern)
  • Lothar Thiele (ETH Zurich)
  • Manuel Lopez-Ibanez (IRIDIA, Code, Brussels)
  • Marco Laumanns (IBM)
  • Robert Synovec (University of Washington, Seattle)
  • Andy McShea (Theo Chocolate)
  • Liz Humston (University of Washington, Seattle)
  • Leonora Bianchi (IDSIA)
  • Will Rowe (Faculty of Life Sciences)
  • Mark Platt (Loughborough)
  • Chris Knight (Faculty of Life Sciences)
  • Philip J. Day (School of Chemistry / MIB)
  • David Wedge (Sanger Institute)
  • Martin Brown (Control Systems Centre, EEE)
  • David Brough (Faculty of Life Sciences)
  • Ben Small (Faculty of Life Sciences)
  • Pedro Mendes (School of Computer Science / MIB)
  • Nancy Rothwell (Faculty of Life Sciences)
  • Marco Dorigo (IRIDIA, Free University of Brussels)
  • Norman Paton (School of Computer Science)
  • Sandra Sampaio (School of Computer Science)
  • Ludi Mikhailov (Manchester Business School)
  • Paul Popelier (School of Chemistry / MIB)
  • Roy Goodacre (School of Chemistry / MIB)
slides