Tomasz Winiarski

Learn More
Many feature selection and feature ranking methods have been proposed. Using real and artificial data an attempt has been made to compare some of these methods. The "feature relevance index" used seems to have little effect on the relative ranking. For continuous features discretization and kernel smoothing are compared. Selection of subsets of features(More)
Many indices for evaluation of features have been considered. Applied to single features they allow for filtering irrelevant attributes. Algorithms for selection of subsets of features also remove redundant features. Hashing techniques enable efficient application of feature relevance indices to selection of feature subsets. A number of such methods have(More)
Feature selection is an essential component in all data mining applications. Ranking of futures was made by several inexpensive methods based on information theory. Accuracy of neural, similarity based and decision tree classifiers calculated with reduced number of features. Comparison with computationally more expensive feature elimination methods was made.
The paper presents a formal approach to robot motion specification. This motion specification takes into account three elementary behaviors that suffice to define any robot interaction with the environment, i.e.: free motion, exerting generalized forces and transition between both of those behaviors. Those behaviors provide a foundation for general motion(More)
This paper describes Rubik's cube manipulation as an example of highly demanding benchmark task for two-handed service robot. Vision is used both for acquiring the cube (servoing) and identifying its initial configuration. During two-handed manipulation force/torque measurements are used to prevent jamming of the cube. Rubik's cube solver, nominal(More)
Two new feature selection methods are introduced, the first based on separability criterion, the second on consistency index that includes interactions between the selected subsets of features. Comparison of accuracy was made against information-theory based selection methods on several data-sets training neurofuzzy and nearest neighbor methods on various(More)
Purpose – This paper seeks to develop universal software (a programming framework) enabling the implementation of service robot controllers. The software should distinguish the hardware-oriented part of the system from the task-oriented one. Moreover, force, vision as well as other sensors should be taken into account. Multi-effector systems have to be(More)
A systematic approach to the design of control systems for robots executing complex tasks will be presented. Each robot is treated as an embodied agent, i.e. an entity influencing a physical environment by its effectors, gathering the information about the state of this environment through its receptors (sensors), and using that information in a decision(More)