Starting with FisPro

contact@fispro.org

Date: March 15, 2018

FisPro (Fuzzy Inference System Professional) allows to create fuzzy inference systems and to use them for reasoning purposes, especially for simulating a physical or biological system. Fuzzy inference systems are briefly described in the fuzzy logic glossary given in the user documentation. They are based on fuzzy rules, which have a good capability for managing progressive phenomenons.

First of all, the FisPro implementation allows to design fuzzy systems from the expert knowledge available in a given field, for instance in winemaking. This approach is illustrated by an example given in the user guide Quickstart with FisPro.

FisPro also allows the complete design of a fuzzy inference system from the numerical data related to the problem under study. Many automatic learning methods unfortunately lead to "black box" systems. In FisPro, constraints are imposed to the algorithms to make the reasoning rules easy to interpret([1]), so that the user understands how the fuzzy system operates. This novel approach is one of the originalities of the software. Some examples are given in the user guide Induction with FisPro.

Both approaches, expert rule design and automatic induction, can be combined to create more complete and better performing systems. FisPro offers educational tools that illustrate the reasoning mechanism, and other tools to measure the system performance on datasets.

This software is made of two distinct parts: a C++ function library, which can be used independently, and a graphical Java interface, which implements most functionalities if the C++ library. It is portable, and can run on most existing platforms.

Authors

• Design and C++ implementation:
  Serge GUILLAUME, Irstea, UMR ITAP
  (http://ser.gui.free.fr/homepage),
  Brigitte CHARNOMORDIC, INRA, UMR MISTEA
  (http://www.inra.fr)

• Java interface : Jean-Luc LABLEE, Irstea, UMR ITAP

• Contributions:
  • C++
    • François OLIVIER, optimization module, from the work by Pierre-Yves GLORENNEC, Algorithmes d'apprentissage pour systèmes d'inférence floue, Edns. Hermes, 1999
    • Sébastien DESTERCKE, orthogonal least squares rule induction (ols)
    • Vincent THERRY, with the support of the Envilys company, cascading systems (superfis)
    • Russel STANDISH, optimized and parallelisable functions (OPENMP standard)
    • Hazaël JONES, design of the implicative systems part
    • Lydie DESPERBEN, C++ implementation of the implicative systems part
  • Java
    • Pierre-Marie BOYER, JNI interface (Java-C++)
    • Mathieu GRELIER, data visualization
    • Anne TIREAU, java interface of the implicative systems part.
  • Communication
    • Jean-Michel FATOU, FisPro icons and web site design
    • Moacir Jr PEDROSO, Portuguese translation of the interface
    • Juan Luis CORTI, Spanish translation of the Quickstart with FisPro manual

Acknowledgments

The development of the first version of FisPro was sustained by public funds from the French government and from Languedoc-Roussillon regional funds, during a research project, COST 2000-012, coordinated by the TRANSFERTS LR association with the industrial partnership of the Cave Cooperative "La Malepère", Arzens, Aude.

FisPro is an open source software, available on the Internet at:

https://www.fispro.org

Elementary notions

A system is also called a FIS, or fuzzy inference system.

The abbreviation MF is used for membership function or fuzzy set (see fuzzy logic glossary, section 3 ).

When starting FisPro, no system is available.

You have a choice between opening an existing system, or creating a new one, either automatically from data, or by hand, element by element.

This introductory guide explains the way to do it in this last case, suited to expert rule definition.

Notes:

• For numerical input, the decimal separator is the point (.).
• Any editing, new input, new output or new MF, is immediately taken in account in the system. Intermediate pop-up windows can be closed without loosing modifications.
• Some options are context dependent. When unavailable, options are grayed out in the menus.
• Expert definition only uses the FIS menu . The Learning menu is for more advanced needs (automatic rule induction).
• The Data menu allows to open an external data file in text format, to visualize data and to perform batch inference.

Note:

The Language option of the Options menu allows to choose the language you want for messages and menus.