Selective Memory Recursive Least Squares (SMRLS) is a recursive least squares (RLS) based real-time training algorithm for linearly parameterized approximators. This project is a demo for radial basis function neural network (RBFNN) based real-time function approximation with SMRLS. The function of each file is explained as follows:
- approximated_function.m -- used to set the function to be approximated by the RBFNN
- RBFNN.m -- return the output of an RBFNN
- centers.m -- used to set the neuron centers of the RBFNN
- SMRLS.m -- SMRLS algorithm
- VDFRLS.m -- variable-direction-forgetting RLS
- FFRLS.m -- RLS with a constant forgetting factor
- SGD.m -- stochastic gradient descent
- randomnurbs.m & RandNurbs.m -- run 'randomnurbs.m' to generate a random non-uniform rational B-splines (NURBS) trajectory
- sinusoidal.m -- used to generate a sinusoidal or spiral trajectory
- main_sim.m -- main simulation program
- plot_sim.m -- used to show the results
- figure_plot.m -- used to generate detailed figures
To use this demo, it is recommended to run the program in the following order: randomnurbs.m/sinusoidal.m → main_sim.m → plot_sim.m