Neuro$-$Fuzzy and Soft Computing Project Outline

Objective:

Develop a neuro$-$fuzzy $($NF$)$ system for engineering design applications. It can be used for system

modeling $($e$.$g$.,$ system state forecasting$),$ pattern classification $($e$.$g$.,$ diagnostics$),$ or intelligent

control. Appropriate learning algorithms should be implemented and applied to train both linear

and nonlinear parameters.

Requirements:

The NF system could have at least THREE inputs, and each input should have at least two membership functions.

Programming should be undertaken in MATLAB environment. The fuzzy reasoning and training should be programmed by your instead of using MATLAB toolboxes.

Hybrid training should be used for parameter optimization: The LSE could be used for linear parameter training and the gradient descent algorithm could be applied for nonlinear system parameter optimization.

The ANFIS toolbox in MATLAB can be used for performance in terms comparison of reasoning and training. The ANFIS model should have similar fuzzy reasoning and training algorithms.

It is recommended to use an error threshold of ${10}^{-5}$ and over $200$ training epochs.

The report should contain sections such as Introduction, Theory Review, Model Construction, Result Analysis, Conclusion, and References.