A machine learning technique called Graph-Based Induction (GBI) efficiently extracts typical patterns from graph data by stepwise pair expansion (pairwise chunking). In this paper, we introduce Graph-Based Induction for general graph structured data, which can handle directed/undirected, colored/uncolored graphs with/without (self) loop and with colored/uncolored links. We show that its time complexity is almost linear with the size of graph. We, further, show that GBI can effectively be applied to the extraction of typical patterns from DNA sequence data and organnochlorine compound data from which to generate classification rules, and that GBI also works as a feature construction component for other machine learning tools.
Pricing goods properly is critical for the further growth of electronic commerce. One price discrimination technique drawn from microeconomics theory has shown promise as regards the trading of information services. This technique, however, has a serious drawback in that it assumes that a seller knows the distribution of buyers’ preferences. Unfortunately, obtaining such data is not always easy. We can incorporate agent technologies into the technique, namely, by gathering sales data and updating information about buyers, and thus improve the performance. However, such an adaptive method can end in failure if each agent becomes self-interested, namely, if the problem of collusion emerges. If collusion does occur, the rationality of any price revision is lost. To solve this problem, we have developed a pricing mechanism that can withstand buyer collusion. We provide a concrete method for calculating quality and price combinations, and then analyze the mechanism theoretically and show its effectiveness.
How to enhance novice learners’ understanding of programs is a major issue in programming education. Our approach to this issue is to provide them with fill-in-blank program problems. A fill-in-blank program problem gives learners a program of which part is blanked out and the program specification. They are required to fill in the blank so that the program specification can be fulfilled. In solving the problem, they need to trace data and control flows of the program. This induces them to think of the processes embedded in the program, enhancing their learning. However, whether learning is enhanced depends on how to make a blank. This paper proposes a method of blanking out an important point of data or control flow of a program to make instructive a fill-in-blank problem. The essence of this method is to find out the important point with program dependence graph in no consideration of semantic aspects of the processes in the program. It can be consequently incorporated into computer-based educational systems. This paper also describes an experiment on the blank-making method with subjects who have experience of programming education. In this experiment, we have ascertained that blanks made by hand follow the blank-making method. The results suggest that it is valid.
In this paper, we propose heuristics using frequency of node usage for speedup of evolutionary learning and verify the utility experimentally for on-line robot behavior learning. Genetic Programming (GP) is an evolutionary way to acquire a program through interaction with an environment. Since behaviors of a robot are described with a program, researches on applying GP to robot behavior learning have been activated. Unfortunately, in most of the studies, the behavior learning is done off-line using simulation, not a real robot. Because convergence of GP is slow, and this makes operation of a real robot quite expensive. However, since situations out of simulation easily happens in a real world, the behavior learning with a real robot (called on-line learning) remains very signifficant. Thus, in order to make on-line behavior learning with GP practical, we propose a novel crossover method for speedup of GP using node usage of a program.