International Journal on Artificial Intelligence Tools

This paper studies various strategies in constrained simulated annealing (CSA), a global optimization algorithm that achieves asymptotic convergence to constrained global minima (CGM) with probability one for solving discrete constrained nonlinear programming problems (NLPs). The algorithm is based on the necessary and sufficient condition for discrete constrained local minima (CLM) in the theory of discrete Lagrange multipliers and its extensions to continuous and mixed-integer constrained NLPs. The strategies studied include adaptive neighborhoods, distributions to control sampling, acceptance probabilities, and cooling schedules. We report much better solutions than the best-known solutions in the literature on two sets of continuous benchmarks and their discretized versions.

Regional Knowledge is useful in identifying patterns of relationships between variables, and it is particularly important in solving constrained global optimization problems. However, regional knowledge is generally unavailable prior to the optimization search. The questions here are: 1) Is it possible for an evolutionary system to learn regional knowledge during the search instead of having to acquire it beforehand? and 2) How can this regional knowledge be used to expedite evolutionary search? This paper defines regional schemata to provide an explicit mechanism to support the acquisition, storage and manipulation of regional knowledge. In a Cultural Algorithm framework, the belief space "contains" a set of these regional schemata, arranged in a hierarchical architecture, to enable the knowledge-based evolutionary system to learn regional knowledge during the search and apply the learned knowledge to guide the search. This mechanism can be used to guide the optimization search in a direct way, by "pruning" the infeasible regions and "promoting" the promising regions. Engineering problems with nonlinear constraints are tested and the results are discussed. It shows that the proposed mechanism is potential to solve complicated non-linear constrained optimization problems, and some other hard problems, e.g. the optimization problems with "ridges" in landscapes.

Recently, there has been much interest in enhancing purely combinatorial formalisms with numerical information. For example, planning formalisms can be enriched by taking resource constraints and probabilistic information into account. The Mixed Integer Programming (MIP) paradigm from operations research provides a natural tool for solving optimization problems that combine such numeric and non-numeric information. The MIP approach relies heavily on linear program relaxations and branch-and-bound search. This is in contrast with depth-first or iterative deepening strategies generally used in artificial intelligence. We provide a detailed characterization of the structure of the underlying search spaces as explored by these search strategies. Our analysis shows that much can be gained by combining different search strategies for solving hard MIP problems, thereby leveraging each strategy's strength in terms of the combinatorial and numeric information.

This article deals with the design of a hybrid learning system. This system integrates different kinds of learning methods and gives a set of class hierarchies as the result. The classes in these hierarchies are very similar. The method occurrences compare their results and automatically refine them to try to make them converge towards a unique hierarchy that unifies all the results. Thus, the system decreases the importance of the initial choices made when initializing classification algorithm (the choice of the method and its parameters) and to solve some of the limitations of the methods used such as an imposed number of classes, non-hierarchical result, or the size of the hierarchy found.

This work deals with stability in incremental induction of decision trees. Stability problems arise when an induction algorithm must revise a decision tree very often and oscillations between similar concepts decrease learning speed. We introduce a heuristic and an algorithm with theoretical and experimental backing to tackle this problem.

We propose a new computational approach to logic-based systems that should reason in a fast but logically sound and complete manner about large-scale complex critical devices and systems that can exhibit unexpected faulty behaviors. The approach is original from at least two points of view. First, it makes use of local search techniques while preserving logical deductive completeness. Second, it proves experimentally efficient for very large knowledge bases thanks to new heuristics in the use of local search techniques.

In this paper, we present a method for improving the performance of classical symbolic rules. This is achieved by introducing a type of hybrid rules, called neurules, which integrate neurocomputing into the symbolic framework of production rules. Neurules are produced by converting existing symbolic rules. Each neurule is considered as an adaline unit, where weights are considered as significance factors. Each significance factor represents the significance of the associated condition in drawing the conclusion. A rule is fired when the corresponding adaline output becomes active. This significantly reduces the size of the rule base and, due to a number of heuristics used in the inference process, increases efficiency of the inferences.

There is growing interest in many application domains for the temporal treatment and manipulation of spatially referenced objects. Handling the time dimension in spatial databases can greatly enhance and extend their functionality and usability by offering means of understanding the spatial behaviour in time. Few works, to date, have been directed towards the development of formalisms for representation and reasoning in this domain. In this paper, a new approach is presented for the representation and reasoning over spatio-temporal relationships. The approach is simple and aims to satisfy the requirements of coherency, expressiveness and reasoning power. Consistent behaviours of spatial objects in time are denoted episodes. The topology of the domain is defined by decomposing episodes into representative components and relationships are defined between those components. Spatio-temporal reasoning is achieved by composing the relationships between the object components using constraint networks. New composition tables between simple spatio-temporal regions and between regions and volumes are also derived and used in the reasoning process.

Knowledge-based engineering and computational intelligence are expected to become core technologies in the design and manufacturing for the next generation of space exploration missions. The literature is contradictory on how we are to assess such systems. Studies indicate significant disagreement regarding the amount of testing needed for system assessment. The sizes of standard black-box test suites are impractically large since the black-box approach neglects the internal structure of knowledge-based systems. On the contrary, practical results repeatedly indicate that only a few tests are needed to sample the range of behaviors of a knowledge-based program.

In this paper, we model testing as a search process over the internal state space of the knowledge-based system. When comparing different test suites, the test suite that examines larger portion of the state space is considered more complete. Our goal is to investigate the trade-off between the completeness criterion and the size of test suites. The results of testing experiment on tens of thousands of mutants of real-world knowledge based systems indicate that a very limited gain in completeness can be achieved through prolonged testing. The use of simple (or random) search strategies for testing appears to be as powerful as testing by more thorough search algorithms.