Building Interpretable Systems in Real Time



Building interpretable learning machines from data, incrementally, with the
capability for a-posteriori knowledge extraction, is still a big challenge. It is difficult to
avoid some degree of redundancy and unnecessary complexity in the obtained models.
This work presents two alternative ways for building interpretable systems, using fuzzy
models or kernel machines. The goal is to control the complexity of the machine, either
the number of rules or the number of kernels, using incremental learning techniques and
merging of fuzzy sets. In the first case, online implementation of mechanisms for
merging membership functions and rule base simplification, toward evolving first-order
Takagi-Sugeno fuzzy systems (eTS), is proposed in this chapter in order to improve the
interpretability of the fuzzy models. As a comparative solution, kernel-based methods
may also contribute to the interpretability issue, particularly the incremental learning
algorithms. The interpretability is analyzed based on equivalence principles between
local model networks and Takagi-Sugeno fuzzy inference systems.
A considerable reduction of redundancy and complexity of the models is obtained,
increasing the model transparency and human interpretability. The learning technique
is developed for large spanned evolving first-order Takagi-Sugeno (eTS) fuzzy models.
The benchmark of the Box-Jenkins time-series prediction is used to evaluate the


computational intelligence

Book Chapter

Evolving Intelligent Systems: Methodology and Applications, 6, pp. 127-150, John Willey and Sons, April 2010

Cited by

Year 2015 : 1 citations

 GSETSK: a generic self-evolving TSK fuzzy neural network with a novel Hebbian-based rule reduction approach
NN Nguyen, WJ Zhou, C Quek - Applied Soft Computing, 2015 - Elsevier
Abstract Takagi–Sugeno–Kang (TSK) fuzzy systems have been widely applied for solving
function approximation and regression-centric problems. Existing dynamic TSK models
proposed in the literature can be broadly classified into two classes. Class I TSK models ...

Year 2013 : 1 citations

 On-line assurance of interpretability criteria in evolving fuzzy systems–achievements, new concepts and open issues
E Lughofer - Information Sciences, 2013 - Elsevier
Abstract In this position paper, we are discussing achievements and open issues in the
interpretability of evolving fuzzy systems (EFS). In addition to pure on-line complexity
reduction approaches, which can be an important direction for increasing the ...

Year 2012 : 1 citations

 Introducing evolving Takagi–Sugeno method based on local least squares support vector machine models
M Komijani, C Lucas, BN Araabi, A Kalhor - Evolving Systems, 2012 - Springer
Abstract In this study, an efficient local online identification method based on the evolving
Takagi–Sugeno least square support vector machine (eTS-LS-SVM) for nonlinear time
series prediction is introduced. As an innovation, this paper has applied the nonlinear ...

Year 2011 : 2 citations

 LIVRO] Evolving fuzzy systems-methodologies, advanced concepts and applications
E Lughofer - 2011 - Springer
In today's industrial systems, economic markets, life and health-care sciences fuzzy systems
play an important role in many application scenarios such as system identification, fault
detection and diagnosis, quality control, instrumentation and control, decision support ..

 Human-inspired evolving machines—the next generation of evolving intelligent systems
E Lughofer - IEEE SMC newsletter, 2011 -
Abstract: In today's real-world applications, there is an increasing need to integrate new
information and knowledge into model-building processes to account for changing system
dynamics, new operating conditions or environmental influences. This is essential to ...