The Number of coefficents of Time delay neural network

4 views (last 30 days)
for the following Time delay neural network
clc; clear all; close all;
[X,T] = simpleseries_dataset;
net1 = timedelaynet(1:2,20);
[Xs,Xi,Ai,Ts] = preparets(net1,X,T);
net1 = train(net1,Xs,Ts,Xi);
y1 = net1(Xs,Xi);
weights1 = getwb(net1)
According to my understanding; the input to this network supposed to be the current input and the previous inputs X(n), X(n-1), X(n-2)
Hence the number of weights supposed to be (3x20 +20x1) and the bias (20+1) , hence the vector od weights and bias suppoed to a vector with length = 101
But, when I use the getwb(net1) I get vector with length = 81 ??!!
why he neglect the weights of one sample

Accepted Answer

Mahesh Taparia
Mahesh Taparia on 4 Feb 2020
It does not neglect any weight. Since the number of input delays is 2, the number of weights will be (2X20+20X1) and the bias (20+1). The vector length will be 81. If the input delay is 3, then it will be 101. For more information you can refer to the documentation page of timedelaynet here.
  1 Comment
Abdelwahab Afifi
Abdelwahab Afifi on 4 Feb 2020
In case of delay =2, then which of the following will the input of the network ?
1- X(n), X(n-1) and X(n-2)
2- or X(n), X(n-1)

Sign in to comment.

More Answers (0)


Find more on Sequence and Numeric Feature Data Workflows in Help Center and File Exchange

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by