<html><!-- Created using the cpp_pretty_printer from the dlib C++ library. See http://dlib.net for updates. --><head><title>dlib C++ Library - krr_classification_ex.cpp</title></head><body bgcolor='white'><pre>
<font color='#009900'>// The contents of this file are in the public domain. See LICENSE_FOR_EXAMPLE_PROGRAMS.txt
</font><font color='#009900'>/*
This is an example illustrating the use of the kernel ridge regression
object from the dlib C++ Library.
This example creates a simple set of data to train on and then shows
you how to use the kernel ridge regression tool to find a good decision
function that can classify examples in our data set.
The data used in this example will be 2 dimensional data and will
come from a distribution where points with a distance less than 13
from the origin are labeled +1 and all other points are labeled
as -1. All together, the dataset will contain 10201 sample points.
*/</font>
<font color='#0000FF'>#include</font> <font color='#5555FF'><</font>iostream<font color='#5555FF'>></font>
<font color='#0000FF'>#include</font> <font color='#5555FF'><</font>dlib<font color='#5555FF'>/</font>svm.h<font color='#5555FF'>></font>
<font color='#0000FF'>using</font> <font color='#0000FF'>namespace</font> std;
<font color='#0000FF'>using</font> <font color='#0000FF'>namespace</font> dlib;
<font color='#0000FF'><u>int</u></font> <b><a name='main'></a>main</b><font face='Lucida Console'>(</font><font face='Lucida Console'>)</font>
<b>{</b>
<font color='#009900'>// This typedef declares a matrix with 2 rows and 1 column. It will be the
</font> <font color='#009900'>// object that contains each of our 2 dimensional samples. (Note that if you wanted
</font> <font color='#009900'>// more than 2 features in this vector you can simply change the 2 to something else.
</font> <font color='#009900'>// Or if you don't know how many features you want until runtime then you can put a 0
</font> <font color='#009900'>// here and use the matrix.set_size() member function)
</font> <font color='#0000FF'>typedef</font> matrix<font color='#5555FF'><</font><font color='#0000FF'><u>double</u></font>, <font color='#979000'>2</font>, <font color='#979000'>1</font><font color='#5555FF'>></font> sample_type;
<font color='#009900'>// This is a typedef for the type of kernel we are going to use in this example.
</font> <font color='#009900'>// In this case I have selected the radial basis kernel that can operate on our
</font> <font color='#009900'>// 2D sample_type objects
</font> <font color='#0000FF'>typedef</font> radial_basis_kernel<font color='#5555FF'><</font>sample_type<font color='#5555FF'>></font> kernel_type;
<font color='#009900'>// Now we make objects to contain our samples and their respective labels.
</font> std::vector<font color='#5555FF'><</font>sample_type<font color='#5555FF'>></font> samples;
std::vector<font color='#5555FF'><</font><font color='#0000FF'><u>double</u></font><font color='#5555FF'>></font> labels;
<font color='#009900'>// Now let's put some data into our samples and labels objects. We do this
</font> <font color='#009900'>// by looping over a bunch of points and labeling them according to their
</font> <font color='#009900'>// distance from the origin.
</font> <font color='#0000FF'>for</font> <font face='Lucida Console'>(</font><font color='#0000FF'><u>double</u></font> r <font color='#5555FF'>=</font> <font color='#5555FF'>-</font><font color='#979000'>20</font>; r <font color='#5555FF'><</font><font color='#5555FF'>=</font> <font color='#979000'>20</font>; r <font color='#5555FF'>+</font><font color='#5555FF'>=</font> <font color='#979000'>0.4</font><font face='Lucida Console'>)</font>
<b>{</b>
<font color='#0000FF'>for</font> <font face='Lucida Console'>(</font><font color='#0000FF'><u>double</u></font> c <font color='#5555FF'>=</font> <font color='#5555FF'>-</font><font color='#979000'>20</font>; c <font color='#5555FF'><</font><font color='#5555FF'>=</font> <font color='#979000'>20</font>; c <font color='#5555FF'>+</font><font color='#5555FF'>=</font> <font color='#979000'>0.4</font><font face='Lucida Console'>)</font>
<b>{</b>
sample_type samp;
<font color='#BB00BB'>samp</font><font face='Lucida Console'>(</font><font color='#979000'>0</font><font face='Lucida Console'>)</font> <font color='#5555FF'>=</font> r;
<font color='#BB00BB'>samp</font><font face='Lucida Console'>(</font><font color='#979000'>1</font><font face='Lucida Console'>)</font> <font color='#5555FF'>=</font> c;
samples.<font color='#BB00BB'>push_back</font><font face='Lucida Console'>(</font>samp<font face='Lucida Console'>)</font>;
<font color='#009900'>// if this point is less than 13 from the origin
</font> <font color='#0000FF'>if</font> <font face='Lucida Console'>(</font><font color='#BB00BB'>sqrt</font><font face='Lucida Console'>(</font><font face='Lucida Console'>(</font><font color='#0000FF'><u>double</u></font><font face='Lucida Console'>)</font>r<font color='#5555FF'>*</font>r <font color='#5555FF'>+</font> c<font color='#5555FF'>*</font>c<font face='Lucida Console'>)</font> <font color='#5555FF'><</font><font color='#5555FF'>=</font> <font color='#979000'>13</font><font face='Lucida Console'>)</font>
labels.<font color='#BB00BB'>push_back</font><font face='Lucida Console'>(</font><font color='#5555FF'>+</font><font color='#979000'>1</font><font face='Lucida Console'>)</font>;
<font color='#0000FF'>else</font>
labels.<font color='#BB00BB'>push_back</font><font face='Lucida Console'>(</font><font color='#5555FF'>-</font><font color='#979000'>1</font><font face='Lucida Console'>)</font>;
<b>}</b>
<b>}</b>
cout <font color='#5555FF'><</font><font color='#5555FF'><</font> "<font color='#CC0000'>samples generated: </font>" <font color='#5555FF'><</font><font color='#5555FF'><</font> samples.<font color='#BB00BB'>size</font><font face='Lucida Console'>(</font><font face='Lucida Console'>)</font> <font color='#5555FF'><</font><font color='#5555FF'><</font> endl;
cout <font color='#5555FF'><</font><font color='#5555FF'><</font> "<font color='#CC0000'> number of +1 samples: </font>" <font color='#5555FF'><</font><font color='#5555FF'><</font> <font color='#BB00BB'>sum</font><font face='Lucida Console'>(</font><font color='#BB00BB'>mat</font><font face='Lucida Console'>(</font>labels<font face='Lucida Console'>)</font> <font color='#5555FF'>></font> <font color='#979000'>0</font><font face='Lucida Console'>)</font> <font color='#5555FF'><</font><font color='#5555FF'><</font> endl;
cout <font color='#5555FF'><</font><font color='#5555FF'><</font> "<font color='#CC0000'> number of -1 samples: </font>" <font color='#5555FF'><</font><font color='#5555FF'><</font> <font color='#BB00BB'>sum</font><font face='Lucida Console'>(</font><font color='#BB00BB'>mat</font><font face='Lucida Console'>(</font>labels<font face='Lucida Console'>)</font> <font color='#5555FF'><</font> <font color='#979000'>0</font><font face='Lucida Console'>)</font> <font color='#5555FF'><</font><font color='#5555FF'><</font> endl;
<font color='#009900'>// Here we normalize all the samples by subtracting their mean and dividing by their standard deviation.
</font> <font color='#009900'>// This is generally a good idea since it often heads off numerical stability problems and also
</font> <font color='#009900'>// prevents one large feature from smothering others. Doing this doesn't matter much in this example
</font> <font color='#009900'>// so I'm just doing this here so you can see an easy way to accomplish this with
</font> <font color='#009900'>// the library.
</font> vector_normalizer<font color='#5555FF'><</font>sample_type<font color='#5555FF'>></font> normalizer;
<font color='#009900'>// let the normalizer learn the mean and standard deviation of the samples
</font> normalizer.<font color='#BB00BB'>train</font><font face='Lucida Console'>(</font>samples<font face='Lucida Console'>)</font>;
<font color='#009900'>// now normalize each sample
</font> <font color='#0000FF'>for</font> <font face='Lucida Console'>(</font><font color='#0000FF'><u>unsigned</u></font> <font color='#0000FF'><u>long</u></font> i <font color='#5555FF'>=</font> <font color='#979000'>0</font>; i <font color='#5555FF'><</font> samples.<font color='#BB00BB'>size</font><font face='Lucida Console'>(</font><font face='Lucida Console'>)</font>; <font color='#5555FF'>+</font><font color='#5555FF'>+</font>i<font face='Lucida Console'>)</font>
samples[i] <font color='#5555FF'>=</font> <font color='#BB00BB'>normalizer</font><font face='Lucida Console'>(</font>samples[i]<font face='Lucida Console'>)</font>;
<font color='#009900'>// here we make an instance of the krr_trainer object that uses our kernel type.
</font> krr_trainer<font color='#5555FF'><</font>kernel_type<font color='#5555FF'>></font> trainer;
<font color='#009900'>// The krr_trainer has the ability to perform leave-one-out cross-validation.
</font> <font color='#009900'>// It does this to automatically determine the regularization parameter. Since
</font> <font color='#009900'>// we are performing classification instead of regression we should be sure to
</font> <font color='#009900'>// call use_classification_loss_for_loo_cv(). This function tells it to measure
</font> <font color='#009900'>// errors in terms of the number of classification mistakes instead of mean squared
</font> <font color='#009900'>// error between decision function output values and labels.
</font> trainer.<font color='#BB00BB'>use_classification_loss_for_loo_cv</font><font face='Lucida Console'>(</font><font face='Lucida Console'>)</font>;
<font color='#009900'>// Now we loop over some different gamma values to see how good they are.
</font> cout <font color='#5555FF'><</font><font color='#5555FF'><</font> "<font color='#CC0000'>\ndoing leave-one-out cross-validation</font>" <font color='#5555FF'><</font><font color='#5555FF'><</font> endl;
<font color='#0000FF'>for</font> <font face='Lucida Console'>(</font><font color='#0000FF'><u>double</u></font> gamma <font color='#5555FF'>=</font> <font color='#979000'>0.000001</font>; gamma <font color='#5555FF'><</font><font color='#5555FF'>=</font> <font color='#979000'>1</font>; gamma <font color='#5555FF'>*</font><font color='#5555FF'>=</font> <font color='#979000'>5</font><font face='Lucida Console'>)</font>
<b>{</b>
<font color='#009900'>// tell the trainer the parameters we want to use
</font> trainer.<font color='#BB00BB'>set_kernel</font><font face='Lucida Console'>(</font><font color='#BB00BB'>kernel_type</font><font face='Lucida Console'>(</font>gamma<font face='Lucida Console'>)</font><font face='Lucida Console'>)</font>;
<font color='#009900'>// loo_values will contain the LOO predictions for each sample. In the case
</font> <font color='#009900'>// of perfect prediction it will end up being a copy of labels.
</font> std::vector<font color='#5555FF'><</font><font color='#0000FF'><u>double</u></font><font color='#5555FF'>></font> loo_values;
trainer.<font color='#BB00BB'>train</font><font face='Lucida Console'>(</font>samples, labels, loo_values<font face='Lucida Console'>)</font>;
<font color='#009900'>// Print gamma and the fraction of samples correctly classified during LOO cross-validation.
</font> <font color='#0000FF'>const</font> <font color='#0000FF'><u>double</u></font> classification_accuracy <font color='#5555FF'>=</font> <font color='#BB00BB'>mean_sign_agreement</font><font face='Lucida Console'>(</font>labels, loo_values<font face='Lucida Console'>)</font>;
cout <font color='#5555FF'><</font><font color='#5555FF'><</font> "<font color='#CC0000'>gamma: </font>" <font color='#5555FF'><</font><font color='#5555FF'><</font> gamma <font color='#5555FF'><</font><font color='#5555FF'><</font> "<font color='#CC0000'> LOO accuracy: </font>" <font color='#5555FF'><</font><font color='#5555FF'><</font> classification_accuracy <font color='#5555FF'><</font><font color='#5555FF'><</font> endl;
<b>}</b>
<font color='#009900'>// From looking at the output of the above loop it turns out that a good value for
</font> <font color='#009900'>// gamma for this problem is 0.000625. So that is what we will use.
</font> trainer.<font color='#BB00BB'>set_kernel</font><font face='Lucida Console'>(</font><font color='#BB00BB'>kernel_type</font><font face='Lucida Console'>(</font><font color='#979000'>0.000625</font><font face='Lucida Console'>)</font><font face='Lucida Console'>)</font>;
<font color='#0000FF'>typedef</font> decision_function<font color='#5555FF'><</font>kernel_type<font color='#5555FF'>></font> dec_funct_type;
<font color='#0000FF'>typedef</font> normalized_function<font color='#5555FF'><</font>dec_funct_type<font color='#5555FF'>></font> funct_type;
<font color='#009900'>// Here we are making an instance of the normalized_function object. This object provides a convenient
</font> <font color='#009900'>// way to store the vector normalization information along with the decision function we are
</font> <font color='#009900'>// going to learn.
</font> funct_type learned_function;
learned_function.normalizer <font color='#5555FF'>=</font> normalizer; <font color='#009900'>// save normalization information
</font> learned_function.function <font color='#5555FF'>=</font> trainer.<font color='#BB00BB'>train</font><font face='Lucida Console'>(</font>samples, labels<font face='Lucida Console'>)</font>; <font color='#009900'>// perform the actual training and save the results
</font>
<font color='#009900'>// print out the number of basis vectors in the resulting decision function
</font> cout <font color='#5555FF'><</font><font color='#5555FF'><</font> "<font color='#CC0000'>\nnumber of basis vectors in our learned_function is </font>"
<font color='#5555FF'><</font><font color='#5555FF'><</font> learned_function.function.basis_vectors.<font color='#BB00BB'>size</font><font face='Lucida Console'>(</font><font face='Lucida Console'>)</font> <font color='#5555FF'><</font><font color='#5555FF'><</font> endl;
<font color='#009900'>// Now let's try this decision_function on some samples we haven't seen before.
</font> <font color='#009900'>// The decision function will return values >= 0 for samples it predicts
</font> <font color='#009900'>// are in the +1 class and numbers < 0 for samples it predicts to be in the -1 class.
</font> sample_type sample;
<font color='#BB00BB'>sample</font><font face='Lucida Console'>(</font><font color='#979000'>0</font><font face='Lucida Console'>)</font> <font color='#5555FF'>=</font> <font color='#979000'>3.123</font>;
<font color='#BB00BB'>sample</font><font face='Lucida Console'>(</font><font color='#979000'>1</font><font face='Lucida Console'>)</font> <font color='#5555FF'>=</font> <font color='#979000'>2</font>;
cout <font color='#5555FF'><</font><font color='#5555FF'><</font> "<font color='#CC0000'>This is a +1 class example, the classifier output is </font>" <font color='#5555FF'><</font><font color='#5555FF'><</font> <font color='#BB00BB'>learned_function</font><font face='Lucida Console'>(</font>sample<font face='Lucida Console'>)</font> <font color='#5555FF'><</font><font color='#5555FF'><</font> endl;
<font color='#BB00BB'>sample</font><font face='Lucida Console'>(</font><font color='#979000'>0</font><font face='Lucida Console'>)</font> <font color='#5555FF'>=</font> <font color='#979000'>3.123</font>;
<font color='#BB00BB'>sample</font><font face='Lucida Console'>(</font><font color='#979000'>1</font><font face='Lucida Console'>)</font> <font color='#5555FF'>=</font> <font color='#979000'>9.3545</font>;
cout <font color='#5555FF'><</font><font color='#5555FF'><</font> "<font color='#CC0000'>This is a +1 class example, the classifier output is </font>" <font color='#5555FF'><</font><font color='#5555FF'><</font> <font color='#BB00BB'>learned_function</font><font face='Lucida Console'>(</font>sample<font face='Lucida Console'>)</font> <font color='#5555FF'><</font><font color='#5555FF'><</font> endl;
<font color='#BB00BB'>sample</font><font face='Lucida Console'>(</font><font color='#979000'>0</font><font face='Lucida Console'>)</font> <font color='#5555FF'>=</font> <font color='#979000'>13.123</font>;
<font color='#BB00BB'>sample</font><font face='Lucida Console'>(</font><font color='#979000'>1</font><font face='Lucida Console'>)</font> <font color='#5555FF'>=</font> <font color='#979000'>9.3545</font>;
cout <font color='#5555FF'><</font><font color='#5555FF'><</font> "<font color='#CC0000'>This is a -1 class example, the classifier output is </font>" <font color='#5555FF'><</font><font color='#5555FF'><</font> <font color='#BB00BB'>learned_function</font><font face='Lucida Console'>(</font>sample<font face='Lucida Console'>)</font> <font color='#5555FF'><</font><font color='#5555FF'><</font> endl;
<font color='#BB00BB'>sample</font><font face='Lucida Console'>(</font><font color='#979000'>0</font><font face='Lucida Console'>)</font> <font color='#5555FF'>=</font> <font color='#979000'>13.123</font>;
<font color='#BB00BB'>sample</font><font face='Lucida Console'>(</font><font color='#979000'>1</font><font face='Lucida Console'>)</font> <font color='#5555FF'>=</font> <font color='#979000'>0</font>;
cout <font color='#5555FF'><</font><font color='#5555FF'><</font> "<font color='#CC0000'>This is a -1 class example, the classifier output is </font>" <font color='#5555FF'><</font><font color='#5555FF'><</font> <font color='#BB00BB'>learned_function</font><font face='Lucida Console'>(</font>sample<font face='Lucida Console'>)</font> <font color='#5555FF'><</font><font color='#5555FF'><</font> endl;
<font color='#009900'>// We can also train a decision function that reports a well conditioned probability
</font> <font color='#009900'>// instead of just a number > 0 for the +1 class and < 0 for the -1 class. An example
</font> <font color='#009900'>// of doing that follows:
</font> <font color='#0000FF'>typedef</font> probabilistic_decision_function<font color='#5555FF'><</font>kernel_type<font color='#5555FF'>></font> probabilistic_funct_type;
<font color='#0000FF'>typedef</font> normalized_function<font color='#5555FF'><</font>probabilistic_funct_type<font color='#5555FF'>></font> pfunct_type;
<font color='#009900'>// The train_probabilistic_decision_function() is going to perform 3-fold cross-validation.
</font> <font color='#009900'>// So it is important that the +1 and -1 samples be distributed uniformly across all the folds.
</font> <font color='#009900'>// calling randomize_samples() will make sure that is the case.
</font> <font color='#BB00BB'>randomize_samples</font><font face='Lucida Console'>(</font>samples, labels<font face='Lucida Console'>)</font>;
pfunct_type learned_pfunct;
learned_pfunct.normalizer <font color='#5555FF'>=</font> normalizer;
learned_pfunct.function <font color='#5555FF'>=</font> <font color='#BB00BB'>train_probabilistic_decision_function</font><font face='Lucida Console'>(</font>trainer, samples, labels, <font color='#979000'>3</font><font face='Lucida Console'>)</font>;
<font color='#009900'>// Now we have a function that returns the probability that a given sample is of the +1 class.
</font>
<font color='#009900'>// print out the number of basis vectors in the resulting decision function.
</font> <font color='#009900'>// (it should be the same as in the one above)
</font> cout <font color='#5555FF'><</font><font color='#5555FF'><</font> "<font color='#CC0000'>\nnumber of basis vectors in our learned_pfunct is </font>"
<font color='#5555FF'><</font><font color='#5555FF'><</font> learned_pfunct.function.decision_funct.basis_vectors.<font color='#BB00BB'>size</font><font face='Lucida Console'>(</font><font face='Lucida Console'>)</font> <font color='#5555FF'><</font><font color='#5555FF'><</font> endl;
<font color='#BB00BB'>sample</font><font face='Lucida Console'>(</font><font color='#979000'>0</font><font face='Lucida Console'>)</font> <font color='#5555FF'>=</font> <font color='#979000'>3.123</font>;
<font color='#BB00BB'>sample</font><font face='Lucida Console'>(</font><font color='#979000'>1</font><font face='Lucida Console'>)</font> <font color='#5555FF'>=</font> <font color='#979000'>2</font>;
cout <font color='#5555FF'><</font><font color='#5555FF'><</font> "<font color='#CC0000'>This +1 class example should have high probability. Its probability is: </font>"
<font color='#5555FF'><</font><font color='#5555FF'><</font> <font color='#BB00BB'>learned_pfunct</font><font face='Lucida Console'>(</font>sample<font face='Lucida Console'>)</font> <font color='#5555FF'><</font><font color='#5555FF'><</font> endl;
<font color='#BB00BB'>sample</font><font face='Lucida Console'>(</font><font color='#979000'>0</font><font face='Lucida Console'>)</font> <font color='#5555FF'>=</font> <font color='#979000'>3.123</font>;
<font color='#BB00BB'>sample</font><font face='Lucida Console'>(</font><font color='#979000'>1</font><font face='Lucida Console'>)</font> <font color='#5555FF'>=</font> <font color='#979000'>9.3545</font>;
cout <font color='#5555FF'><</font><font color='#5555FF'><</font> "<font color='#CC0000'>This +1 class example should have high probability. Its probability is: </font>"
<font color='#5555FF'><</font><font color='#5555FF'><</font> <font color='#BB00BB'>learned_pfunct</font><font face='Lucida Console'>(</font>sample<font face='Lucida Console'>)</font> <font color='#5555FF'><</font><font color='#5555FF'><</font> endl;
<font color='#BB00BB'>sample</font><font face='Lucida Console'>(</font><font color='#979000'>0</font><font face='Lucida Console'>)</font> <font color='#5555FF'>=</font> <font color='#979000'>13.123</font>;
<font color='#BB00BB'>sample</font><font face='Lucida Console'>(</font><font color='#979000'>1</font><font face='Lucida Console'>)</font> <font color='#5555FF'>=</font> <font color='#979000'>9.3545</font>;
cout <font color='#5555FF'><</font><font color='#5555FF'><</font> "<font color='#CC0000'>This -1 class example should have low probability. Its probability is: </font>"
<font color='#5555FF'><</font><font color='#5555FF'><</font> <font color='#BB00BB'>learned_pfunct</font><font face='Lucida Console'>(</font>sample<font face='Lucida Console'>)</font> <font color='#5555FF'><</font><font color='#5555FF'><</font> endl;
<font color='#BB00BB'>sample</font><font face='Lucida Console'>(</font><font color='#979000'>0</font><font face='Lucida Console'>)</font> <font color='#5555FF'>=</font> <font color='#979000'>13.123</font>;
<font color='#BB00BB'>sample</font><font face='Lucida Console'>(</font><font color='#979000'>1</font><font face='Lucida Console'>)</font> <font color='#5555FF'>=</font> <font color='#979000'>0</font>;
cout <font color='#5555FF'><</font><font color='#5555FF'><</font> "<font color='#CC0000'>This -1 class example should have low probability. Its probability is: </font>"
<font color='#5555FF'><</font><font color='#5555FF'><</font> <font color='#BB00BB'>learned_pfunct</font><font face='Lucida Console'>(</font>sample<font face='Lucida Console'>)</font> <font color='#5555FF'><</font><font color='#5555FF'><</font> endl;
<font color='#009900'>// Another thing that is worth knowing is that just about everything in dlib is serializable.
</font> <font color='#009900'>// So for example, you can save the learned_pfunct object to disk and recall it later like so:
</font> <font color='#BB00BB'>serialize</font><font face='Lucida Console'>(</font>"<font color='#CC0000'>saved_function.dat</font>"<font face='Lucida Console'>)</font> <font color='#5555FF'><</font><font color='#5555FF'><</font> learned_pfunct;
<font color='#009900'>// Now let's open that file back up and load the function object it contains.
</font> <font color='#BB00BB'>deserialize</font><font face='Lucida Console'>(</font>"<font color='#CC0000'>saved_function.dat</font>"<font face='Lucida Console'>)</font> <font color='#5555FF'>></font><font color='#5555FF'>></font> learned_pfunct;
<b>}</b>
</pre></body></html>