If we knew what it was we were doing, it would not be called research, would it?

Albert Einstein

Research interests

My current research interests cover a broad area of intelligent and biologically/ nature inspired learning and adaptive systems and include a wide range of machine learning and hybrid intelligent techniques encompassing data and information fusion, learning and adaptation methods, multiple classifier and prediction systems, processing and modelling of uncertainty in pattern recognition, diagnostic analysis and decision support systems.

My personal and our group's past, current and near future activities concentrate within the following 3 overlapping areas:

• Nature inspired and smart adaptive systems - focusing on investigations into novel techniques and inspirations coming from nature in order to move a step closer towards truly smart and adaptive systems. Apart from various biologically inspired approaches to computationally intelligent systems, one of the very distinctive areas of our work is related to the Physics of Information research stream where the metaphors between physical particles and data have been used to develop novel pattern classification, clustering and data condensation techniques.

• Data and information fusion including multiple classifier and prediction systems - focusing on combination of different sources of data and their modalities; issues concerning the uncertainty processing related to producing predictions which are generated with reliable confidence limits; investigations of multilevel organisations of classifiers and predictors which from our recent, first of this kind, theoretical results based on derivation of majority vote limits for systems structured in this way have huge potential.

• Hybrid intelligent systems and their applications - focusing on hybridization and/or combination of different well established and novel intelligent technologies which are frequently more beneficial to be applied in combination rather than exclusively. This stream is the longest running initiative within my personal research interests with the initial developments and inspirations coming from the application of computational intelligence techniques to the problem of pattern based fault diagnosis in water distribution systems characterised by large-scale non-linear models and noisy and frequently incomplete measurements.

Here are few bullet points with more catchy names which give a fuller picture of my research interests and activities.

 

General areas Computational intelligence and soft computing Data and information fusion Pattern classification and clustering Information theory Machine learning Mathematical modelling, simulation and control theory Smart adaptive systems

Application areas Decision support Optimisation Pattern recognition and classification Condition monitoring Fault diagnosis Time series prediction Data mining Data visualization Intelligent information retrieval

Some more specific research interest areas related to the current and past research projects

Nature-inspired techniques - artificial neural networks, evolutionary computation including genetic algorithms and evolutionary strategies, fuzzy systems, neuro-fuzzy systems, hybrid intelligent systems, particle swarm optimization, ant colonies, simulated annealing, physically inspired computational intelligence approaches.

Hybrid neuro-fuzzy clustering and classification systems based on hyperbox fuzzy sets - combination of supervised and unsupervised learning; growing structure; missing data; partially labelled data; on-line and hierarchical learning algorithms; pruning procedures; data editing.

Classification and prediction models generated directly from data using highly flexible non-parametric methods - algorithm independent learning approaches; resampling techniques; error estimation; model selection techniques; model validation; control of model complexity.

Data and information fusion - data, feature, decision and multilevel fusion; adaptive and self improving fusion systems architectures; fusion learning in imperfect, imprecise and incomplete environments; intelligent techniques for fusion processing.

Multiple classifier systems and combination of predictions- combination methods; diversity measures; resampling techniques; performance estimation; ensemble component selection and generation methods. 

Diagnostic analysis - novelty detection; condition monitoring; pattern based fault detection and location; imprecise and incomplete data processing.

Optimisation - use of various optimization criteria for large scale non-linear systems; design and application of analogue NNs for state estimation and confidence limit analysis; evolutionary algorithms in multicriteria optimization problems.

Mathematical modelling and simulation - visualization techniques; modelling of dynamical systems; partial differential equations; time series analysis; modelling and simulation of distribution systems.