Reduced Precision Bayesian Network Classifiers

Seminar Type: - None -
Student: Carlos

 Short Description

Bayesian Network Classifiers (BNCs) are probabilistic classifiers that can be applied in a wide range of applications, e.g. digit recognition, medical expert systems or speech recognition. While these classifiers are theoretically well understood, there are little results on real-world implementations available. Aim of this project is to close this gap. This includes:

  • Theory:
    • Literature review
    • Learning of BNCs that are suited for implementation in hardware, i.e. definition and implementation of appropriate learning algorithms
    • Evaluation of the proposed learning algorithms
    • If possible, derivation of suboptimality bounds on these BNCs in comparison to "optimal" BNCs
  • Practice:
    • Implementation of BNCs on FPGAs and comparison of different implementation options
    • Evaluation of the performance of the implemented system
    • Usage of BNCs in a real world application, e.g. handwritten digit recognition

Illustration of a BNC with naive Bayes structure

Your Profile/Requirements

The candidate should be interested in machine learning, applied mathematics/statistics, Matlab programming, and algorithms. Interested candidates are encouraged to ask for further information. Additionally, the supervision of own projects in one of the above mention fields is possible.


Franz Pernkopf ( or 0316/873 4436)


Sebastian Tschiatschek, Peter ReinprechtManfred MückeFranz Pernkopf: Bayesian Network Classifiers with Reduced Precision Parameters. ECML/PKDD (1) 2012: 74-89;