User Tools

Site Tools



  1. Introduction
    1. problem statement, motivation
    2. precise definitions and terminology (define terms)
      1. Pattern, Attribute, Feature/Predictor
    3. define one goal and list consequent objectives
  2. Stateof-the-art
    1. Overview
    2. Inductive Logic Programming
    3. Statistical relational learning
    4. SVM, Tree, Neural, Distance based
    5. Graph Mining
    6. Multi view
    7. Propositionalisation
  3. Objectives
    1. The goal is to get an automated solution into industry. Hurdles: money & time.
      1. Saves time, takes little time
      2. Simple or well known mechanism (SQL) → minimal training
      3. Compatible with current processes (do not fix what ain't broken) → minimal training
      4. Scalable (SQL can be executed)
      5. Simple deployment (no need to install databases…)
      6. Well documented
      7. Make it viral: easily extensible (like evolutionary algorithm)
      8. Ability to recover from a failure and continue from the last stored position
      9. React intelligently to addition/removal of columns (updates) (test first the new columns)
      10. React intelligently to change of columns (begin from the previously best patterns & attributes)
      11. Fast deployment execution, slow analytical phase
      12. Minimize data transfers (allows remote work, better from security point)
      13. Computation with several targets in mind
      14. Focus just on one thing (otherwise it would be tough to stay on schedule)
  4. Contribution
    1. approach,
    2. implementation
    3. experimentation
      1. describe selection of databases (# of references, # columns, #rows, field)
      2. describe selection of comparison algorithms (draw graph with references)
      3. describe HW and SW for time comparison
      4. describe accuracy evaluation (Wilcoxon test)
    4. users feedback from projects
  5. Conclusions
    1. dissadvantage: only univariate evaluation → delegation to classifier
  6. bibliografie, seznam obrazku, tabulek, seznam patentu, publikovanych clanku, list of abbreviations