---
myst:
    substitutions:
        bloc:
            python: 6-22
            cpp: 6-22
---
# The Vectorial Quantification method

This method is called when instead of having a list of input parameters (in terms of stochastic 
distribution) that one would like to transform into a {{doe}}, the user has a dataset, made out of a 
very large number of points. In this case, it is possible, using the `TNeuralGas` class, to create 
sub-sample of points that would be representative of the complete provided-set, based on a NeuralGas 
algorithm. This might be useful in order to test the output of long and complicated codes or 
computations without leaving aside a possible area of the input parameter values.

{numref}`sampler_vq_plot` shows the effect of the reduction of the sample in the simple case of a 
two-dimensional plane, when considering the "geyser.dat" file and its sub-sample of 50 points.

{{
    "```{" "figure" "} " + parent_dir +
        "/roottest/build/uranie/doc/sampler/bloc_codes_chapter/cpp/producePlotsSampler/NeuralGas_reduction.png\n"
    ":align: center\n"
    ":name: sampler_vq_plot\n"
    + figure_scale + "\n"
    "\n"
    "Example of a dataset reduction (the geyser one) using the NeuralGas algorithm, to go from 272 
    points (left) to 50 one (right)\n"
    "```"
}}

Here is an example of how to use the neuralgas algorithm to reduce a database.

{{
    "```{literalinclude} " + parent_dir
    + "/roottest/uranie/doc/sampler/bloc_codes_chapter/" + language + "/vectorial_quantification."
    + extension
    + "\n"
    + ":language: " + language + "\n"
    + ":lines: " + bloc[language] + "\n"
    + "```"
}}

**Construction of the plot**

1. Creation of `TCanvas` subdivided into two pieces to compare results.

2. Construction of the neuralgas object from the provided database

3. Get the sub-sample as a new dataserver, after looping 20 the algorithm

4. Access the latest histogram drawn on current pad, to change its title