Decomposable (probabilistic) models are log-linear models generated by acyclic hypergraphs, and a number of nice properties enjoyed by them are known. In many applications the following selection problem naturally arises: given a probability distribution $p$ over a finite set $V$ of $n$ discrete variables and a positive integer $k$, find a decomposable model with tree-width $k$ that best fits $p$. If $\mathscr{H}$ is the generating hypergraph of a decomposable model and $p_{\mathscr{H}}$ is the estimate of $p$ under the model, we can measure...

We present a test for identifying clusters in high dimensional data based on the k-means algorithm when the null hypothesis is spherical normal. We show that projection techniques used for evaluating validity of clusters may be misleading for such data. In particular, we demonstrate that increasingly well-separated clusters are identified as the dimensionality increases, when no such clusters exist. Furthermore, in a case of true bimodality, increasing the dimensionality makes identifying the correct...

The objective of this paper is to develop feasible gait patterns that could be used to control a real hexapod walking robot. These gaits should enable the fastest movement that is possible with the given robot's mechanics and drives on a flat terrain. Biological inspirations are commonly used in the design of walking robots and their control algorithms. However, legged robots differ significantly from their biological counterparts. Hence we believe that gait patterns should be learned using the...

The problem of extracting more compact rules from a rule-based knowledge base is approached by means of a chunking mechanism implemented via a neural system. Taking advantage of the parallel processing potentialities of neural systems, the computational problem normally arising when introducing chuncking processes is overcome. Also the memory saturation effect is coped with using some sort of forgetting mechanism which allows the system to eliminate previously stored, but less often used chunks....

Designing classifiers may follow different goals. Which goal to prefer among others depends on the given cost situation and the class distribution. For example, a classifier designed for best accuracy in terms of misclassifications may fail when the cost of misclassification of one class is much higher than that of the other. This paper presents a decision-theoretic extension to make fuzzy rule generation cost-sensitive. Furthermore, it will be shown how interpretability aspects and the costs of...

This paper presents a new learning algorithm for the design of Mamdani- type or fully-linguistic fuzzy controllers based on available input-output data. It relies on the use of a previously introduced parametrized defuzzification strategy. The learning scheme is supported by an investigated property of the defuzzification method. In addition, the algorithm is tested by considering a typical non-linear function that has been adopted in a number of published research articles. The test stresses on...

Nowadays, multiclassifier systems (MCSs) are being widely applied in various machine learning problems and in many different domains. Over the last two decades, a variety of ensemble systems have been developed, but there is still room for improvement. This paper focuses on developing competence and interclass cross-competence measures which can be applied as a method for classifiers combination. The cross-competence measure allows an ensemble to harness pieces of information obtained from incompetent...

This paper details nonlinear Model-based Predictive Control (MPC) algorithms for MIMO processes modelled by means of neural networks of a feedforward structure. Two general MPC techniques are considered: the one with Nonlinear Optimisation (MPC-NO) and the one with Nonlinear Prediction and Linearisation (MPC-NPL). In the first case a nonlinear optimisation problem is solved in real time on-line. In order to reduce the computational burden, in the second case a neural model of the process is used...

First, a fuzzy system based on ifFirst, a fuzzy system based on if-then rules and with parametric consequences is recalled. Then, it is shown that the globalthen rules and with parametric consequences is recalled. Then, it is shown that the global and local ε-insensitive learning of the above fuzzy system may be presented as a combination of both an ε-insensitive gradient method and solving a system of linear inequalities. Examples are given of using the introduced method to design fuzzy models...

We discuss a prescriptive approach to multistage optimal fuzzy control of a fuzzy system, given by a fuzzy state transition equation. Fuzzy constraints and fuzzy goals at consecutive control stages are given, and their confluence, Bellman and Zadeh's fuzzy decision, is an explicit performance function to be optimized. First, we briefly survey previous basic solution methods of dynamic programming (Baldwin and Pilsworth, 1982) and branch-and-bound (Kacprzyk, 1979), which are plagued by low numerical...

Assessing the number of clusters of a statistical population is one of the essential issues of unsupervised learning. Given n independent observations X1,...,Xn drawn from an unknown multivariate probability density f, we propose a new approach to estimate the number of connected components, or clusters, of the t-level set $ℒ\left(t\right)=\{x:f(x)\ge t\}$. The basic idea is to form a rough skeleton of the set $ℒ\left(t\right)$ using any preliminary estimator of f, and to count the number of connected components of the resulting graph. Under...

This paper describes a heuristic forecasting model based on neural networks for stock decision-making. Some heuristic strategies are presented for enhancing the learning capability of neural networks and obtaining better trading performance. The China Shanghai Composite Index is used as case study. The forecasting model can forecast the buying and selling signs according to the result of neural network prediction. Results are compared with a benchmark buy-and-hold strategy. The forecasting model...

The problem of developing a general methodology for system design has always been demanding. For this purpose, an evolutionary algorithm, adapted with design-specific representation data structures is devised. The representation modeling the system to be designed, is composed of three levels of abstraction: the first, is an 'abstract brain' layer - mainly a number of competing finite state machines, which in turn control the second level composed of fuzzy Petri nets; the third level constitutes...

Motion control of a mobile manipulator is discussed. The objective is to allow the end-effector to track a given trajectory in a fixed world frame. The motion of the platform and that of the manipulator are coordinated by a neural network which is a kind of graph designed from the kinematic model of the system. A learning paradigm is used to produce the required reference variables for each of the mobile platform and the robot manipulator for an overall coordinate behavior. Simulation results are...

The aim of this paper is to show that the theory of (generalized) random systems with complete connection may serve as a mathematical framework for learning and adaption. Chapter 1 is of an introductory nature and gives a general description of the problems with which one is faced. In Chapter 2 the mathematical model and some results about it are explained. Chapter 3 deals with special learning and adaption models.

This paper presents a methodology for developing fuzzy knowledge based systems (KBS), which permits a complete automatization. This methodology will be useful for approaching more complex problems that those in which machine learning from examples are successful.

The similarity judgerment of two sequences is often decomposed in similarity judgements of the sequence events with an alignment process. However, in some domains like speech or music, sequences have an internal structure which is important for intelligent processing like similarity judgements. In an alignment task, this structure can be reflected more appropriately by using two levels instead of aligning event by event. This idea is related to the structural alignment framework by Markman and Gentner....