Mining software specifications is an important research topic, since a software specification is essential for program testing and verification to ensure the correctness of a software. Lo et at. proposed a software specification miner, called BOB, which finds recurrent rules from a sequence database. A recurrent rule has a suitable property to represent temporal notions such as “eventuality” and “regularity” (i.e., always). In BOB, the mining task is realized, among others, by two steps: first generating rule pre/post-conditions and then forming rules by pairing them. This paper proposes a new method, Interleaved Bidirectional Recurrent Rule Miner (iBiRM for short); the idea of the iBiRM algorithm is to interleave the two steps so that intermediate data in the preceding algorithm can be eliminated. It also incorporates a new support counting method tailored to recurrent rule mining to reduce the number of scanning a given database. We also give experimental results, which show that our algorithm improves on the previous approach.

QuantMiner, proposed by Salleb-Aouissi et al., is one of the well-known systems for mining quantitative association rules using a genetic algorithm (GA). We have applied the GA-based methods of QuantMiner to multi-relational data mining (MRDM), where mining rules involves multiple relations from a relational database, and our preliminary experiments showed that a straightforward application of QuantMiner to MRDM has an efficiency problem. In this paper, we propose RelQM-J, a Java implementation of QuantMiner applied to multi-relational databases (MRDBs), with an aim to handle the computation of the supports of rules in an efficient way, by using a hash-based data structure. Our experimental results show the effectiveness of the proposed method, compared with the conventional method used in QuantMiner.

No abstract available.

Abstract We consider some extensions of co-logic programming and study its relationship with the Horn μ -calculus by Charatonik et al. We first considernegation elimination (NE), a familiar technique of program transformation, for co-logic programs. Given a programP, NE derives itsdualprogram P* which defines the “complement” ofP. When we apply NE to co-logic programs with negation, we show that the stratification restriction, a syntactic condition imposed on co-logic programs, becomes too restrictive in general, and that the Horn μ -calculus can be used as an extension of co-logic programming for handling “non-stratified” co-logic programs. We then consider some applications of non-stratified co-logic programs to the well-founded semantics (WFS) and Answer Set Programming. In particular, we give new iterated fixpoint characterizations of the WFS as well as answer sets via dual programs. We also discuss some applications of non-stratified co-logic programs to program transformation such as partial deduction, and a proof procedure for the WFS.

Implements a display system for cat's cradle game animation in 3D. The input to our system, finger actions, originates from a user. Based on this input, we animate the finger actions and display an operated pattern of string figures. A stabilizing procedure is in charge of obtaining a stable state of the string figures. Processing string figures on a computer is difficult, since the string's dynamics are still unknown. Our system uses a set-of-crossings representation method for processing string figures. We employ some local transformations of the string figures to stabilize them. Furthermore, topological characteristics of the string figures can be observed.

We characterize the problem of detecting edges in images as a fuzzy reasoning problem. The edge detection problem is divided into three stages: filtering, detection, and tracing. Images are filtered by applying fuzzy reasoning based on local pixel characteristics to control the degree of Gaussian smoothing. Filtered images are then subjected to a simple edge detection algorithm which evaluates the edge fuzzy membership value for each pixel, based on local image characteristics. Finally, pixels having high edge membership are traced and assembled into structures, again using fuzzy reasoning to guide the tracing process. The filtering, detection, and tracing algorithms are tested on several test images. Comparison is made with a standard edge detection technique.