Tasks 6 and 5 of a university 'pattern recognition' basic course.
Qt plotting widget - QCustomPlot.
-
Model a random sample of n-dimension objects.
Modelled sample objects are distributed normally, the mean and the variance have to be predefined in a configuration file.
-
Plot objects' 2d-projections.
A number of each dimension is defined in an application form (starting with 0).
-
Classify objects using chosen algorithm (Parzen windows, k nearest neighbours, Bayes classifier, Parametric classifier).
-
Calculate a classification error.
-
Calculate a transformation matrix.
An element tij of this matrix stands for the number of elements of a class j classified as a class j.
You'll need a test.txt to define the number of classes, the number of dimensions, prior probabilities, for each class its distribution parameters. Using QtCreator open lab6.pro, run the project. Input the size of a training sample in a textbox signed N train, the size of a test sample in a N test textbox, click Model to model the sample, click test to test a chosen classifier.
-
Model a random sample of n-dimension objects.
Modelled sample objects are distributed normally, the mean and the variance have to be predefined in a configuration file.
-
Plot objects' 2d-projections.
A number of each dimension is defined in an application form (starting with 0).
-
Classify objects to minimize risk and to minimize error.
-
Calculate an average risk and a classification error.
-
Calculate a transformation matrix.
An element tij of this matrix stands for the number of elements of a class j classified as a class j.
You'll need a config.txt to define the number of classes, the number of dimensions, loss matrix, prior probabilities, for each class its distribution parameters. Using QtCreator open lab5.pro, run the project. Input the size of a sample in a textbox signed N, click Model.
The first plot is min error classification, the second is modelled data, the third plot is min risk classification, colors represent classes.
