Chair of Artificial Intelligence and Machine Learning
print
Breadcrumb Navigation
Content

Fuzzy Pattern Trees as Deep Fuzzy Systems (Ba/Ma)

Topic for a bachelor/master's thesis

Short Description:

The notion of fuzzy pattern tree (FPT) refers to a class of hierarchical models, in which basic properties of data objects are represented in the form of fuzzy sets and successively combined into more complex properties by means of generalized aggregation functions [1]. Different machine learning algorithms have been proposed for constructing such models in a data-driven way, either using a top-down or a bottom-up strategy [3]. FPTs can be used for both classification and regression tasks and have a number of appealing properties, especially compared to traditional “flat” fuzzy models [4].

The goal of this thesis is to revisit FPTs from the perspective of deep learning, i.e., to view pattern trees as a specific type of deep model. In particular, there is an interest in understanding how the depth of an FPT influences important properties such as predictive accuracy. To this end, existing learning algorithms for FPTs and their implementation [2] must be extended, and suitable experimental studies must be conducted. These studies should also include a comparison with deep neural networks. PREREQUISITES: Good background in machine learning, especially supervised learning (e.g., classification, regression), deep learning, and empirical performance studies; algorithm design; programming skills; basic knowledge in fuzzy logic (advantageous but not a prerequisite).

Prerequisites

Good background in machine learning, especially supervised learning (e.g., classification, regression), deep learning, and empirical performance studies; algorithm design; programming skills; basic knowledge in fuzzy logic (advantageous but not a prerequisite).

Contact

Prof. Eyke Hüllermeier

References