Speaker: Prof. Wladyslaw Homenda,Warsaw University of Technology
Time&Place：2019. 6. 20, 9:00—11:30, 后主楼1129;
2019. 6. 21, 9:00—11:30, 后主楼1129；
2019. 6. 23, 9:00—11:30, 后主楼1129；
2019. 6. 24, 9:00—11:30, 后主楼1129；
2019. 6. 25, 9:00—11:30, 后主楼1129；
2019. 6. 26, 9:00—11:30, 后主楼1129.
Title：Rejecting Option in Pattern Recognition Problem
Abstract：
Pattern recognition has established itself as an advanced area with a well-defined methodology, a plethora of algorithms, and well-defined application areas. Prudently formulated evaluation strategies of methods of pattern recognition, especially pattern classifiers, constitute the crux of numerous pattern classifiers.
With the abundance of data, its volume, and existing diversity arises an evident challenge that needs to be carefully addressed. It deals with the quality perceived in a very general sense and manifesting in numerous ways. Missing data, data affected by noise, of foreign patterns, limited precision, information granularity, imbalanced data are commonly encountered phenomena one has to take into consideration in building pattern classifiers and carrying out data analysis. In particular, one has to engage suitable ways of transforming (preprocessing) data (patterns) prior to their analysis, classification, and interpretation.
Dealing with unknown data impacts the very essence of pattern recognition and calls for thorough investigations of the principles of the area and exhibits a direct impact on architectures and development schemes of the classifiers. The proposed course is intended to cover the essentials of pattern recognition and casts in a new perspective of data - in essence we advocate that a new framework of pattern recognition along with its methodology and algorithms has to be established to cope with the challenges of data coming from unknown sources and unavailable at the stage of classifier construction. Such data are called foreign data in contrast to native data produced represented by standard learning sets available for classifier constructor. This is the essential objective of the course.
In what follows a detailed (albeit tentative) outline is presented, focal points are identified as well as elaborate on the main topics to be covered in the consecutive lectures. In general, the approach adhered in the exposure of the material to is the one of top-down form: lectures start with fundamentals, move on to methodological issues and afterwards concentrate on detailed algorithms and applications.
-------------------------------------------------
The Agenda
1. Pattern recognition problem, feature space construction.
2. Basic classifiers: k-nn, decision trees, random forests, SVMs, naive Bayes.
3. Rejecting option of pattern recognition: problem formulation, basic concepts, related ideas.
4. Rejecting option of pattern recognition: architectures of classifiers.
5. Evaluating pattern recognition without and with rejection.
6. Pattern recognition with rejection: empirical analysis.
