一种具有核密度估计和带宽选择的无参数经验贝叶斯分类器的matlab实现.zip

wkiro ===== An implementation of a parameterless empiric Bayesian classifier with a kernel estimator and bandwidth selection. Fully implemented in Matlab, partially in JS. Usage instructions --- `main.m` is the application entry point. The simplest way to run the whole thing is to just type `main` in Matlab's Command Window. The application execution can be parameterized by the following variables: - `TRAINING_SET_PARAMETERS`, specifying a matrix of parameter vectors for each training dataset. - Generated sets can be adjusted in the following dimensions: - X,Y coordinates of the training set center, - radius of the training set. - The number of parameter vectors in the matrix also specifies the number of classes, to which the target elements will be ascribed. - One parameter vector describes the features of one training set. - The data format for the training set parameter matrix is the following: ```matlab [x1,y1,r1; x2,y2,r2; x3,y3,r3] % unrestricted matrix length ``` - In case no value is defined for the variable, the following default is used: ```matlab [-15,-15,1; 15,15,1; -15,15,1; 15,-15,1]; ``` - `TRAINING_SET_ELEMENTS_COUNT`, specifying the quantity of elements forming each training set. - Every training set has an equal number of elements. - In case no value is defined for the variable, the following default is used: ```matlab 20 ``` - `ELEMENTS`, specifying the set of elements to classify. - An element is represented by a pair of coordinates. - The format for the data within this variable is as follows: ```matlab [x1,y1; x2,y2; x3,y3] % unrestricted matrix length ``` - In case no value is defined for the variable, a set of 100 elements generated randomly within a circular area cenetered at `(0,0)` and with a radius of `7` is used as the default. - `KERNEL`, specifying the function used for kernel estimation by the classifier. - The variable is meant to be a function handle and can be used only as such. - There are a few predefined kernel functions available in the `kernels` directory for you to use. - In case no value is defined for the variable, the Gaussian kernel is used as the default. todo --- - update the implementation in JS - add unit test coverage in JS - fix the bug under Octave.