Dr. John A. Bullinaria

School of Computer Science
University of Birmingham
Birmingham
B15 2TT
UK

j.bullinaria @ physics.org

I started my second early retirement in September 2019. Prior to that I was a Senior Lecturer in the School of Computer Science at the University of Birmingham, and I still retain an honorary position there. My academic career began as a Theoretical Physicist with a PhD on supergravity and other unified field theories from the University of Southampton, followed by a post-doctoral research position in the Mathematics Department of Durham University working on superstring theory and quantum gravity. I then took a very early retirement to "travel the world". Having seen enough of the world (i.e., run out of money), I returned to academia three years later by retraining in Artificial Intelligence. I then took up a series of research fellowships in the Psychology departments of The University of Edinburgh, Birkbeck College London and the University of Reading working on various computational modelling projects. I switched to Computer Science and moved to the University of Birmingham in 2001.

My current research interests are mainly in the fields of Computational Intelligence, Cognitive Science, and Artificial Life, particularly those aspects involving Neural and Evolutionary Computation. Major projects in the past have involved models of brain damage (connectionist neuropsychology), language processing (reading, spelling, past tense production, lexical decision), adaptive control (particularly oculomotor control), the optimization of neural information processing architectures (including the emergence of modularity), and the formulation of more biologically realistic evolutionary computation algorithms. Recently I have been mainly working on simulating the evolution of neural systems: exploring the emergence of modularity, the optimization of learning algorithms and learning strategies, critical periods for learning, the interaction of learning and evolution, aspects of Life History Evolution, and models of strategies for coping with changing environments. I have also worked on evolutionary computation approaches to real-world optimization applications such as vehicle routing and bin packing, and continue to work on developing and testing corpus derived semantic representations.

Links to the rest of this web site

Publications page - contains a full list of my peer-reviewd publications, with most of them available to download.

Brief CV - find out which ten universities have been lucky enough to have me, when I was there, and what I did.

Teaching and Academic Admin pages - links to modules I have taught and associated lecture notes/handouts, PhD students I have supervised, and details of admin roles carried out.

Research Interests

My current research interests cover a number of inter-related areas:

  • Artificial Life - agent-based simulations, the interaction of lifetime learning and evolution, meme-based models of learning, learning and evolutionary strategies in changing environments, and aspects of life history evolution.
  • Evolutionary computation - biologically inspired algorithms, the evolution of efficient neural network systems, the Baldwin Effect, individual differences, variable neural plasticity, incremental learning, and the evolution of modularity.
  • Representations of lexical semantics (working with Joe Levy, University of Roehampton) - their extraction from large spoken and written corpora, their optimization, their validation, and their use in models of natural language processing.
  • Operations Research - real-world optimization applications such as vehicle routing problems and bin packing.
  • Adaptable motor/sensor control systems - traditional and neural network models, optimization by learning and evolution, the interaction of learning and evolution, models of oculomotor control, and applications to robotics.
  • Connectionist neuropsychology - the simulation of brain damage using artificial neural networks, the implications for the inference from double dissociation to modularity, and the problems of small scale artefacts in such simulations.
  • Models of reading, spelling and past tense acquisition - NETtalk style models that do not require pre-processing of the training data, the incorporation of semantic routes, and the simulation of developmental and acquired dyslexias.
  • Models of lexical decision - managing without an explicit lexicon, cascaded activation approaches to modeling reaction times, and the simulation of semantic, associative and morphological priming.
  • Neural network performance with noisy and ambiguous data - mixture models, the multi-target approach, and strategies for coping with missing context information.
  • Understanding the internal representations of trained neural networks - analysis of hidden unit activations using principal component analysis, hierarchical cluster analysis, multi-dimensional scaling, discriminant analysis, and output weight projections.
  • Computer-aided gambling - time series prediction, systems based on statistics, neural computation and evolutionary computation.

    Links to other things I am/was associated with:

    Corpus Derived Semantic Representations - The outputs of my research with Joe Levy on lexical semantics.

    Workshop on Distributional Methods in Language Modelling which took place at the University of Birmingham on 28 August 2002.

    The Neural Computation and Psychology Workshop (NCPW) Series which I regularly contribute to.

    This page is maintained by John Bullinaria. Last updated on 10 January 2024.