Podcast: Play in new window | Download | Embed
A feedforward radial basis function network architecture.
The Radial Basis Function (RBF) network is a more advanced machine learning algorithm which is capable of learning and creating novel features to support pattern recognition. Running and interpreting the program is similar to the Linear Machine Software Described in Episode 13 but there is an additional option which specifies the “number of hidden units”. In the figure below, we have 2 targets and 3 hidden units. The RBF network tries to learn a set of hidden units which is capable of helping the RBF make better predictions. Try out the RBF with different numbers of hidden units on different data sets! The “formulas” for the RBF are almost the same as the Linear Machine software but they involve “hidden features”. Here are the formulas for the case where we have 2 targets, 3 hidden units, 4 input features, and 1 intercept.
Equations specifying the RBF multilayer learning machine network architecture.
Important Download Information
- Please remember that you must have a password to download the software. Obtain a password by joining the Learning Machines 101 community by visiting: www.learningmachines101.com
- You will also need a program such as WINZIP or 7-ZIP to unzip the software.
- Please remember that you must download and install the MCR Library before you can use this windows executable software!
- If you have MATLAB already installed on your computer (Windows, MAC, LINUX, or UNIX), then you can still run this program using the MATLAB source code and you do not have to download the MCR Library.
- Please visit Episode 13 and the Linear Machine Software Download page for more specific details regarding download procedures!!!
Windows Executable Code
This is Windows executable code. In order for this to execute you must have already installed the MATLAB Compiler Run-Time Library Installer for Windows Version 2011b. If you have already installed this library, then it is not necessary to re-install the library on your machine. This software is a compiled version of the MATLAB Source Code listed below.
Download Nonlinear Machine Windows Executable Software Here!
MAC OS-X Executable Code
This is MAC OS-X executable code. In order for this to execute you must have already installed the MATLAB Compiler Run-Time Library Installer for MAC OS-X Version 2012a. If you have already installed this library, then it is not necessary to re-install the library on your machine. This software is a compiled version of the MATLAB Source Code listed below.
Also please note that if you are using a very recent version of MAC OS-X you may have to download Xquartz from http://xquartz.macosforge.org/landing/
Download Nonlinear Machine MAC OS-X Software Here!
MATLAB Source Code
The MATLAB Linear Machine software requires that you have installed MATLAB on your computer which is a software development environment that can be purchased from: www.mathworks.com. If MATLAB is not installed on your machine, however, you might still want to download the MATLAB software since it provides a technical specification of the Linear Machine Windows and MAC OS-X Executable code as a collection of text files.
