Unified theories of cognition
Outlier detection for high dimensional data
SIGMOD '01 Proceedings of the 2001 ACM SIGMOD international conference on Management of data
Machine Learning
Constructing rough mereological granules of classifying rules and classifying algorithms
Technologies for constructing intelligent systems
Feature Selection Using Multi-Objective Genetic Algorithms for Handwritten Digit Recognition
ICPR '02 Proceedings of the 16 th International Conference on Pattern Recognition (ICPR'02) Volume 1 - Volume 1
Ontologies: A Silver Bullet for Knowledge Management and Electronic Commerce
Ontologies: A Silver Bullet for Knowledge Management and Electronic Commerce
A Survey of Outlier Detection Methodologies
Artificial Intelligence Review
Toward knowledge-rich data mining
Data Mining and Knowledge Discovery
Data Mining and Knowledge Discovery
Rough sets in perception-based computing
PReMI'05 Proceedings of the First international conference on Pattern Recognition and Machine Intelligence
Hi-index | 0.00 |
Domain, or background, knowledge has proven to be a key component in the development of high-performance classification systems, especially when the objects of interest exhibit complex internal structures, as in the case of images, time series data or action plans. This knowledge usually comes in extrinsic forms such as human expert advices, often contains complex concepts expressed in quasi-natural descriptive languages and need to be assimilated by the classification system. This paper presents a framework for the assimilation of such knowledge, equivalent to matching different ontologies of complex concepts, using rough mereology theory and rough set methods. We show how this framework allows a learning system to acquire complex, highly structured concepts from an external expert in an intuitive and fully interactive manner. We also argue the needs to focus on expert's knowledge elicited from outlier or novel samples, which we deem have a crucial impact on the classification process. Experiment results show that the proposed methods work well on a large collection of handwritten digits, though they are by no means limited to this particular type of data.