FIRNet Demo 1 -- The Henon Attractor

FIRNet Demo 1 -- The Henon Attractor

Input data set has 5000 real vectors
FIR network was trained on all 5000 and used to predict an extra 1500
Network had 2 layers with 1:7:1 nodes and 4:3 taps
After 150 epochs with a learning rate of 0.007, MSE was 0.000057
Data can be plotted against itself with a one-step delay to get a 2-D orbit

Henon Attractor data set -- first 150 points

Henon Attractor Orbit -- 5000 points

Henon Attractor Orbit -- 1500 predicted points

The FIRNet Page

Forward to Demo #2