ANFIS APPLICATION
By Justin J. Padinjaremury
Background: ANFIS Layers
Layer 1: Contains the membership functions with
adaptive parameters
Layer 2: Relationship between membership functions
with fixed parameters (calculates firing strength)
Layer 3: Normal
ME697
Assignment #1
Due: 1/26
1. Using fuzzy sets A = cfw_0.3/ x1 0.4 / x2 and B = cfw_0.1 / y1 0.5 / y2
Spring 2012
0.3 / y3 , calculate the
Cartesian production of A B and the cylindrical extension of A B to set B.
2. A fuzzy system with one input and
ME697
Assignment #2
Due: 2/23/10
Spring 2012
1. Construct a three layer feedforward neural network (one hidden layer) using the back
propagation learning algorithm to approximate the following two input and one output
function: (10 pts)
2
2
2
2
1
x1
x13
OrthogonalLeastSquaresTraining
Orthogonal
AlgorithmforRadialBasisFunction
Networks
Jie Cai
02/12/2012
Radial Basis Function Networks(RBFN)
( x ci
RBF:
)
e.g., thin-plate-spline function:
( ) = 2 log ( )
1
Gaussian RBF:
(
p x|j
2
(
p x|j
)
n
)
x m j 2
Neural Network Training
LINEAR LEAST SQUARES METHOD
Nathan Toner
Outline
Brief overview of the method
2-dimensional falling object dynamics
Network setup
Simulation results
Conclusions
Method
Train radial basis function network (RBFN) using linear least-s