Adaptive Modelling, Estimation and Fusion from Data

1. An introduction to modelling and learning algorithms.- 1.1 Introduction to modelling.- 1.2 Modelling, control and learning algorithms.- 1.3 The learning problem.- 1.4 Book philosophy and contents overview.- 1.4.1 Book overview.- 1.4.2 A historical perspective of adaptive modelling and control.- 2. Basic concepts of data-based modelling.- 2.1 Introduction.- 2.2 State-space models versus input-output models.- 2.2.1 Conversion of state-space models to input-output models.- 2.2.2 Conversion of input-output models to state-space models.- 2.3 Nonlinear modelling by basis function expansion.- 2.4 Model parameter estimation.- 2.5 Model quality.- 2.5.1 The bias-variance dilemma.- 2.5.2 Bias-variance balance by model structure regularisation.- 2.6 Reproducing kernels and regularisation networks.- 2.7 Model selection methods.- 2.7.1 Model selection criteria.- 2.7.2 Model selection criteria sensitivity.- 2.7.3 Correlation tests.- 2.8 An example: time series modelling.- 3. Learning laws for linear-in-the-parameters networks.- 3.1 Introduction to learning.- 3.2 Error or performance surfaces.- 3.3 Batch learning laws.- 3.3.1 General learning laws.- 3.3.2 Gradient descent algorithms.- 3.4 Instantaneous learning laws.- 3.4.1 Least mean squares learning.- 3.4.2 Normalised least mean squares learning.- 3.4.3 NLMS weight convergence.- 3.4.4 Recursive least squares estimation.- 3.5 Gradient noise and normalised condition numbers.- 3.6 Adaptive learning rates.- 4. Fuzzy and neurofuzzy modelling.- 4.1 Introduction to fuzzy and neurofuzzy systems.- 4.2 Fuzzy systems.- 4.2.1 Fuzzy sets.- 4.2.2 Fuzzy operators.- 4.2.3 Fuzzy relation surfaces.- 4.2.4 Inferencing.- 4.2.5 Fuzzification and defuzzification.- 4.3 Functional mapping and neurofuzzy models.- 4.4 Takagi-Sugeno local neurofuzzy model.- 4.5 Neurofuzzy modelling examples.- 4.5.1 Thermistor modelling.- 4.5.2 Time series modelling.- 5. Parsimonious neurofuzzy modelling.- 5.1 Iterative construction modelling.- 5.2 Additive neurofuzzy modelling algorithms.- 5.3 Adaptive spline modelling algorithm (ASMOD).- 5.3.1 ASMOD refinements.- 5.3.2 Illustrative examples of.- 5.4 Extended additive neurofuzzy models.- 5.4.1 Weight identification.- 5.4.2 Extended additive model structure identification.- 5.5 Hierarchical neurofuzzy models.- 5.6 Regularised neurofuzzy models.- 5.6.1 Bayesian regularisation.- 5.6.2 Error bars.- 5.6.3 Priors for neurofuzzy models.- 5.6.4 Local regularised neurofuzzy models.- 5.7 Complexity reduction through orthogonal least squares.- 5.8 A-optimality neurofuzzy model construction (NeuDec).- 6. Local neurofuzzy modelling.- 6.1 Introduction.- 6.2 Local orthogonal partitioning algorithms.- 6.2.1 k-d Trees.- 6.2.2 Quad-trees.- 6.3 Operating point dependent neurofuzzy models.- 6.4 State space representations of operating point dependent neurofuzzy models.- 6.5 Mixture of experts modelling.- 6.6 Multi-input-Multi-output (MIMO) modelling via input variable selection.- 6.6.1 MIMO NARX neurofuzzy model decomposition.- 6.6.2 Feedforward Gram-Schmidt OLS procedure for linear systems.- 6.6.3 Input variable selection via the modified Gram-Schmidt OLS for piecewise linear submodels.- 7. Delaunay input space partitioning modelling.- 7.1 Introduction.- 7.2 Delaunay triangulation of the input space.- 7.3 Delaunay input space partitioning for locally linear models.- 7.4 The Bezier-Bernstein modelling network.- 7.4.1 Neurofuzzy modelling using Bezier-Bernstein function for univariate term fi(xi) and bivariate term fi1, j1(xi1, xj1).- 7.4.2 The complete Bezier-Bernstein model construction algorithm.- 7.4.3 Numerical examples.- 8. Neurofuzzy linearisation modelling for nonlinear state estimation.- 8.1 Introduction to linearisation modelling.- 8.2 Neurofuzzy local linearisation and the MASMOD algorithm.- 8.3 A hybrid learning scheme combining MASMOD and EM algorithms for neurofuzzy local linearisation.- 8.4 Neurofuzzy feedback linearisation (NFFL).- 8.5 Formulation of neurofuzzy state estimators.- 8.6 An example of nonlinear trajectory estimation.- 9. Multisensor data fusion using Kaiman filters based on neurofuzzy linearisation.- 9.1 Introduction.- 9.2 Measurement fusion.- 9.2.1 Outputs augmented fusion (OAF).- 9.2.2 Optimal weighting measurement fusion (OWMF).- 9.2.3 On functional equivalence of OAF and.- 9.2.4 On the decentralised architecture.- 9.3 State-vector fusion.- 9.3.1 State-vector assimilation fusion (SVAF).- 9.3.2 Track-to-track fusion (TTF).- 9.3.3 On the decentralised architecture.- 9.4 Hierarchical multisensor data fusion - trade-off between centralised and decentralised Architectures.- 9.5 Simulation examples.- 9.5.1 On functional equivalence of two measurement fusion methods.- 9.5.2 On hierarchical multisensor data fusion.- 10. Support vector neurofuzzy models.- 10.1 Introduction.- 10.2 Support vector machines.- 10.2.1 Loss functions.- 10.2.2 Feature space and kernel functions.- 10.3 Support vector regression.- 10.4 Support vector neurofuzzy networks.- 10.5 SUPANOVA.- 10.6 A comparison among neural network models.- 10.7 Conclusions.- References.