Document Actions
Course: Concept lattices for (practical) data analysis and knowledge processing
This is a 3-day course on formal methods (Formal Concept Analysis) for data analysis, to be held April 17th-19th.
This is a master course on formal methods (Formal Concept Analysis) for data analysis. The lecturer is Sergei Kuznetsov, Head of the School of Applied Mathematics and Information Science (more info on the author).
The schedule of the course can be found here. You are all welcome to attend it.
Time: Apr 17th-19th, 15:00-18:00
Place: A4204, Campus Nord, Barcelona.
ABSTRACT:
Methods for analyzing data based on Formal Concept Analysis (FCA) [18, 6] fit
well to the paradigm of knowledge discovery, which “is interactive and iterative,
involving numerous steps with many decision being made by the user” [7]. How-
ever, the stages of knowledge discovery process where data are automatically
analyzed involve application of methods of machine learning or data analysis. In
our course we would like to relate FCA to mathematical models of data analysis
and machine learning, which underlie methods of knowledge discovery.
On the one hand, techniques related to extraction of knowledge from data,
like attribute exploration [6] and generation of implication bases [2] and partial
implication bases [14], were in the mainstream of FCA research from the very
beginning. On the other hand, certain well-known methods of data analysis and
machine learning, such as clustering and biclustering, induction of decision trees,
generation of version spaces, JSM-method and generating bases of association
rules are naturally expressed in terms of FCA, which helps their understanding
and realization.
SCHEDULE:
The schedule of the course can be found here. You are all welcome to attend it.
Time: Apr 17th-19th, 15:00-18:00
Place: A4204, Campus Nord, Barcelona.
ABSTRACT:
Methods for analyzing data based on Formal Concept Analysis (FCA) [18, 6] fit
well to the paradigm of knowledge discovery, which “is interactive and iterative,
involving numerous steps with many decision being made by the user” [7]. How-
ever, the stages of knowledge discovery process where data are automatically
analyzed involve application of methods of machine learning or data analysis. In
our course we would like to relate FCA to mathematical models of data analysis
and machine learning, which underlie methods of knowledge discovery.
On the one hand, techniques related to extraction of knowledge from data,
like attribute exploration [6] and generation of implication bases [2] and partial
implication bases [14], were in the mainstream of FCA research from the very
beginning. On the other hand, certain well-known methods of data analysis and
machine learning, such as clustering and biclustering, induction of decision trees,
generation of version spaces, JSM-method and generating bases of association
rules are naturally expressed in terms of FCA, which helps their understanding
and realization.
SCHEDULE:
Concept lattices for (practical) data analysis and knowledge processing
1. Introduction: concept lattices for applications (1 hour)
2. Baisc notions: relations, orders, lattices and their graphs (1 hour)
3. Introduction to FCA: contexts, concepts, concept lattice, implications (2 hours)
4. Algorithm for constructing all concepts and lattice diagram (2 hours)
5. Implications, functional dependencies, bases, Attribute Exploration (2 hours)
6. Learning with concept lattices: association rules, hypotheses, version spaces, decision trees, biclustering (2 hours)
7. Ontology engineering with FCA: constructing taxonomies, selecting best concepts, finding nonvertical relations. (2 hours)
