Skip to content

Radial Basis Function(RBF) network implementation using Evolution Strategy(ES) to train

Notifications You must be signed in to change notification settings

AradAshrafi/RBF-network-implementation-using-ES-to-train

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

20 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

functionality of this program :

  • Single-Layer Radial Basis Function (RBF) Architecture Implementation

    X contains inputs
    X =[ (X1)   ...  (XL)]
    G is matrix which contains gi(Xj) (which are outputs of each RBF node)
    gi = e (-‫γ‬i (X - Vi)^T(X-Vi))       ‫γi‬ is ith constant and Vi is ith center
    G = [ g1(X1)   ...   gm(X1)
            .
            .
            .
            g1(XL) ... gm(XL) ]
    W is a vector contains weights between RBF layer and output
    W =[ (w1)
            .
            .
            .
            (wm) ]
    ‫̂y‬= GW

  • Single-Layer RBF weights calculation

    W = (GT G)-1 GT y

  • Single-Layer RBF error calculation

    L is our loss function (to calculate error)
    ‫‪L(ŷ,y) = 1/2 - (ŷ - y)T(ŷ - y)

  • Evolution Strategy algorithm with V(vector) and ‫‫‫‫‫‫‫γ(scalar) parameters

    • ES(MU, LAMBDA)
    • Mutation strategy needs to be more precise
    • alternative chromosome length
  • Reach High accuracy for regression

    • Show Results with current Architecture for Regression Problem

    regression result

  • Reach High accuracy for two class

    • Show Results with current Architecture for Classification Problem binary classification result
  • Reach High accuracy for (multi class) classification

    • loss function need to be more accurate

    classification result

About

Radial Basis Function(RBF) network implementation using Evolution Strategy(ES) to train

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages