Learning from High-Dimensional Data (PhD Thesis)

The author is currently available for talks and to teach classes on the subjects (contact the author for this end).
Please use this bib file (and associated publications) to cite the work.

The contents of the thesis can be accessed below (compact format):

  
   I-II. Foundations 65 pages (available)

I.  Introduction
    Abstract, Acknowledgments and Introduction
    Universe of Discourse
    Problem Motivation
    Solution Space and Contents Organization
II. Assessing Models Learned from High-Dimensional Data
    Performance Guarantees of Classification Models
    Performance Guarantees of Local Descriptive Models
    Synthetic Data Generation for Robust Assessments

   

  
   III. Learning Local Descriptive Models from Tabular Data 151 pages (available)

C1 A Structured View on Pattern-based Biclustering
C2 Biclustering Robust to Noise, Missings and Discretization Problems
C3 Additive, Multiplicative and Symmetric Models for Tabular Data
C4 Flexible Plaid Models: Meaningful Interactions between Biclusters
C5 Scalable Pattern-based Biclustering
C6 Flexible and Robust Order-Preserving Biclustering
C7 Biclustering with Efficient Closing Options
C8 Effective Biclustering of Large-Scale Network Data
C9 Flexible Pattern-based Biclustering: Putting All Together
C10 Constraint-based Biclustering with Domain Knowledge

   

  
   IV. Learning Local Descriptive Models from Structured Data 62 pages (provided via request, soon available here)

C1 Learning Cascade Models from Multivariate Time Series
C2 Learning Local Descriptive Models from Multi-sets of Events
C3 Stochastic Modeling of Itemset Sequences
C4 Advanced Stochastic Modeling of Structured Data

   

  
   V. Significance Guarantees of Local Descriptive Models 42 pages (available)

C1 Assessing the Statistical Significance of Biclustering Solutions
C2 Significance of Biclusters with Flexible Coherencies
C3 Assessing the Significance of Real-valued Biclusters
C4 Significance of Local Descriptive Models from Structured Data

   

  
   VI. Learning Effective Associative Classifiers 88 pages (available)

C1 Effective Associative Classification from Discriminative Biclusters
C2 Classification from Regions with Non-Constant Coherency
C3 Advanced Aspects of Associative Classification
C4 Learning Associative Classifiers from Structured Data
C5 Classification from Statistically Significant Regions
C6 Learning Significant and Accurate Decisions
C7 Multi-period Classification for Predictive Tasks

   

  
   VII. Conclusions, Future Work and Bibliography 32 pages (available)

Validation, Contributions and Implications
Future Work
References

   

For an introductory view of the covered contents in this dissertation:

  
   Synthesis of Contents and Contributions 35 pages (available)   

-->