Extracting rules from neural networks using symbolic algorithms: preliminary results

Milaré, Claudia Regina; Carvalho, André C. P. L. F. de; Monard, Maria Carolina

doi:10.1109/iccima.2001.970500

Cited by 4 publications

(5 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Pedagogical methods include decision trees-based proxy models considering the network as a black box and using input data and the network's predictions as targets to learn the proxies. The standard approach considers as targets the hard labels given by the network, i.e., discrete classes [21,78,63]. Distillation methods, however, consider soft labels as targets, which are the probabilities predicted by the network [19,30,55].…”

Section: Global Methodsmentioning

confidence: 99%

A Systematic Literature Review of Machine Learning Applications for Port's Operations

Mekkaoui

Benabbou

Berrado

2020

2020 5th International Conference on Logistics Operations Management (GOL)

View full text Add to dashboard Cite

Motivated by the interpretability question in ML models as a crucial element for the successful deployment of AI systems, this paper focuses on rule extraction as a means for neural networks interpretability. Through a systematic literature review, different approaches for extracting rules from feedforward neural networks, an important block in deep learning models, are identified and explored. The findings reveal a range of methods developed for over two decades, mostly suitable for shallow neural networks, with recent developments to meet deep learning models' challenges. Rules offer a transparent and intuitive means of explaining neural networks, making this study a comprehensive introduction for researchers interested in the field. While the study specifically addresses feedforward networks with supervised learning and crisp rules, future work can extend to other network types, machine learning methods, and fuzzy rule extraction.

show abstract

Section: Global Methodsmentioning

confidence: 99%

A Systematic Literature Review of Machine Learning Applications for Port's Operations

Mekkaoui

Benabbou

Berrado

2020

2020 5th International Conference on Logistics Operations Management (GOL)

View full text Add to dashboard Cite

show abstract

“…Field tests [7] have demonstrated that the excitation system damper is very effective in controlling torsional interaction and operating experience has been satisfactory. Since then, a lot of papers have been published about excitation system dampers [8][9][10][11][12][13][14]. Most of these control schemes failed to stabilize all the torsional modes for the all series compensation ratios.…”

Section: Introductionmentioning

confidence: 99%

“…Using 25 fuzzy if-then rules have increased the complexity and thereby computational cost, too. Seo et al [13] designed a LQG based PSS to damp the SSR modes. Parameter uncertainties are considered in the controller design stage and only the measurable signals are used; however the simulations are only for the 50% series compensation ratio.…”

Section: Introductionmentioning

confidence: 99%

Damping subsynchronous resonance using a novel excitation controller

Ghorbani

Hasanzadeh

Pourmohammad

et al. 2009

2009 IEEE International Conference on Industrial Technology

View full text Add to dashboard Cite

A novel supplementary subsynchronous damping controller (SSDC) is proposed for the generator excitation system which is capable of damping out subsynchronous oscillations in power systems with series compensated transmission lines. The offered SSDC uses a filtered feedback from generator acceleration power signal and after passing through a proportional gain it is added to automatic voltage regulator (AVR) output. Two filters are used and the cut off frequencies are specified intelligently in order to affect only the modes that are critical to damping capability of the system. The IEEE Second Benchmark problem is considered as the studying system. The simulation results proved that the designed controller can be successfully used in damping the oscillations for the all series compensation ratios even when small and large disturbances influence the system. Simplicity, practicality and high performance of the offered controller are other advantages of the proposed controller.

show abstract

“…Although rule extraction for modelling process is a very valuable problem in artificial intelligence applications, extracting fuzzy model of the plant using a systematic approach is not always easily solvable. Some methods for model extraction are discussed in [12][13][14]. Recently, fuzzy Takagi-Sugeno-Kang models have attracted much attention [15].…”

Section: Introductionmentioning

confidence: 99%

Extracting fuzzy If-Then rules using a neural network identifier with application to Boiler-Turbine system

Pourmohammad¹,

Afzalian²

2009

2009 IEEE International Conference on Control Applications

View full text Add to dashboard Cite

In this paper a feed-forward neural network is proposed to extract Fuzzy Hyperbolic Model (FHM) of industrial plants. FHMs resemble Takagi-Sugeno-Kang (TSK) fuzzy models in general, however have some advantages. FHM is an inherently nonlinear model and can capture all the nonlinearities of the system. On the other hand there are some systematic approaches to design and analysis such models. The synergy between Artificial Neural Networks (ANN), which are notorious for their black-box character, and fuzzy logic proved to be particularly successful. Such a synergy allows combining the powerful learning-from-examples capability of ANNs with the high-level symbolic information processing of fuzzy logic systems. The offered network is used to obtain the parameters of the plant from input-output data. It is shown that there is a unique transformation from the proposed network to hyperbolic model of the plant and vice versa. Parameters of the fuzzy model can be obtained from weights and biases in trained network. Boiler-Turbine system is considered as a case study to show how the proposed ANN can be used to extract the fuzzy model. The obtained model is validated by some input-output data provided from the reference model. Simulation results proved the effectiveness of the offered neural network in extracting the fuzzy model of the plant. 978-1-4244-4602-5/09/$25.00 ©2009 IEEE

show abstract

Extracting rules from neural networks using symbolic algorithms: preliminary results

Cited by 4 publications

References 13 publications

A Systematic Literature Review of Machine Learning Applications for Port's Operations

A Systematic Literature Review of Machine Learning Applications for Port's Operations

Damping subsynchronous resonance using a novel excitation controller

Extracting fuzzy If-Then rules using a neural network identifier with application to Boiler-Turbine system

Contact Info

Product

Resources

About