We propose two general heuristics to transform a batch Hillclimbing search into an incremental one. Then, we apply our heuristics to two Bayesian network structure learning algorithms and experimentally see that our incremental approach saves a significant amount of computing time while it yields similar networks than the batch algorithms.

Many learning systems search through a space of possible performance elements, seeking an element with high expected utility. As the task of nding the globally optimal element is usually intractable, many practical learning systems use hill-climbing to nd a local optimum. Unfortunately, even this is diicult, as it depends on the distribution of problems, which is typically unknown. This paper a...

| Delta coding is an iterative genetic search strategy that sustains search by periodically re-initializing the population. This helps to avoid premature convergence during genetic search. Delta coding also remaps hyper-space with each iteration in an attempt to locate \easier" search spaces with respect to genetic search. Here, the optimization ability of delta coding is compared against the C...

Enforced hill-climbing is an effective deterministic hillclimbing technique that deals with local optima using breadth-first search (a process called “basin flooding”). We propose and evaluate a stochastic generalization of enforced hill-climbing for online use in goal-oriented probabilistic planning problems. We assume a provided heuristic function estimating expected cost to the goal with fla...

OCR post-processing is an important step to improve text accuracy. It includes two main tasks, error detection and correction. Hill climbing algorithm a heuristic search method used for solving optimization problems. In this paper, we present novel correction approach using adapted version of the algorithm. Correction candidates errors are explored by random character edits evolved with climbin...

We describe an extensive study of search in GSAT, an approximation procedure for propositional satissability. GSAT performs greedy hill-climbing on the number of satissed clauses in a truth assignment. Our experiments provide a more complete picture of GSAT's search than previous accounts. We describe in detail the two phases of search: rapid hill-climbing followed by a long plateau search. We ...

This study presents an investigation of enhancing the capability of the Scatter Search (SS) metaheuristic in guiding the search effectively toward elite solutions. Generally, SS generates a population of random initial solutions and systematically selects a set of diverse and elite solutions as a reference set for guiding the search. The work focuses on three strategies that may have an impact ...

This paper presents a method for pattern discovery based on viewpoints and feature set patterns. The representation for pattern components accommodates in a fully general way the taxonomic relationships that may exist between interval classes. A heuristic probabilistic hill climbing algorithm is developed to rapidly direct the search towards interesting patterns. The method can be used for sing...