What do we learn from binding features? Evidence for multilevel feature integration.

Four experiments were conducted to investigate the relationship between the binding of visual features (as measured by their aftereffects on subsequent binding) and the learning of feature-conjunction probabilities. Both binding and learning effects were obtained, but they did not interact. Interestingly, (shape-color) binding effects disappeared with increasing practice, presumably because of the fact that only 1 of the features involved was relevant to the task. However, this instability was only observed for arbitrary, not highly overlearned combinations of simple geometric features and not for real objects (colored pictures of a banana and strawberry), where binding effects were strong and resistant to practice. These findings suggest that learning has no direct impact on the strength or resistance of bindings or on speed with which features are bound; however, learning does affect the amount of attention particular feature dimensions attract, which again can influence which features are considered in binding. (PsycINFO Database Record (c) 2011 APA, all rights reserved)     

The Design of Discrimination Experiments

Experimentation plays a fundamental role in scientific discovery. Scientists experiment to gather data, investigate phenomena, measure quantities, and test theories. In this article, we address the problem of designing experiments to discriminate between two completing theories. Given an initial situation for which the two theories make the same prediction, the experiment design problem is to determine how to modify the situation such that the two theories make different predictions for the modified situation. The modified situation is called a discrimination experiment. We present a knowledge-intensive method called DEED for designing discrimination experiments. The method analyzes the differences in the two theories' explanations of the prediction for the initial situation. Based on this analysis, it determines modifications to the initial situation that will result in a discrimination experiment. We illustrate the method with the design of experiments to discriminate between several pairs of qualitative theories in the fluids domain.     

A machine learning system for identifying transmembrane domains from amino acid sequences

We present our machine learning system, that uses inductive logic programming techniques to learn how to identify transmembrane domains from amino acid sequences. Our system facilitates the use of operators such as ‘contains’, that act on entire sequences, rather than on individual elements of a sequence. The prediction accuracy of our new system is around 93%, and this compares favourably with earlier results. This work was carried out with the support of a research grant from ISIS, Fujitsu Laboratories.     

MKR──一种有效的增量式概念获取系统

本文对示例式学习的ＩＤ５Ｒ算法进行了深入的研究并作了进一步的改进，提高了其计算效率．并提出一种适合于增量式学习的抗干扰机制，拓广了算法的适用范围．     

用奇异模型描述的受限机器人的学习控制

本文了一类受限机器人的迭代学习控制方案，控制器的设计是基于机器人模型的降阶形式。在存在有界未知干扰的情况下，对末端操纵器受线性、无摩擦约束面的受限机器人。本文给出的控制方案保证了机器人系统的完全运动跟踪，同时保证了力跟踪误差有界的，且界的大小是可调节的。     

Real-World Robotics: Learning to Plan for Robust Execution

In executing classical plans in the real world, small discrepancies between a planner's internal representations and the real world are unavoidable. These can conspire to cause real-world failures even though the planner is sound and, therefore, proves that a sequence of actions achieves the goal. Permissive planning, a machine learning extension to classical planning, is one response to this difficulty. This paper describes the permissive planning approach and presents GRASPER, a permissive planning robotic system that learns to robustly pick up novel objects.     

Successive pattern classification based on test feature classifier and its application to defect image classification

A novel successive learning algorithm based on a Test Feature Classifier is proposed for efficient handling of sequentially provided training data. The fundamental characteristics of the successive learning are considered. In the learning, after recognition of a set of unknown data by a classifier, they are fed into the classifier in order to obtain a modified performance. An efficient algorithm is proposed for the incremental definition of prime tests which are irreducible combinations of features and capable of classifying training patterns into correct classes. Four strategies for addition of training patterns are investigated with respect to their precision and performance using real pattern data. A real-world problem of classification of defects on wafer images has been dealt with by the proposed classifier, obtaining excellent performance even through efficient addition strategies.     

Post-Supervised Template Induction for Information Extraction from Lists and Tables in Dynamic Web Sources

Dynamic web sites commonly return information in the form of lists and tables. Although hand crafting an extraction program for a specific template is time-consuming but straightforward, it is desirable to automatically generate template extraction programs from examples of lists and tables in html documents. Supervised approaches have been shown to achieve high accuracy, but they require manual labelling of training examples, which is also time consuming. Fully unsupervised approaches, which extract rows and columns by detecting regularities in the data, cannot provide sufficient accuracy for practical domains. We describe a novel technique, Post-supervised Learning, which exploits unsupervised learning to avoid the need for training examples, while minimally involving the user to achieve high accuracy. We have developed unsupervised algorithms to extract the number of rows and adopted a dynamic programming algorithm for extracting columns. Our method achieves high performance with minimal user input compared to fully supervised techniques.