Plotters¶
The plotters are the building blocks of BEAT that provide the visualization for the output result of BEAT experiments, such as bar plots or ROC curves.
Definition of a Plotter¶
A plotter just like other building blocks of BEAT consists of two main components:
a JSON object with several fields, specifying the inputs, the produced figure, the parameters and additional information such as the language in which it is implemented.
source code describing how to produce the desired figure using the given parameters. They are mostly implemented in python and use matplotlib library.
JSON Declaration¶
The JSON declaration of plotters consists of several field. Here is an example of a plotter that produces bar plots:
{
"dataformat": "plot/bar/1",
"description": "Basic bar plotter for simple histograms",
"language": "python",
"parameters": {
"axis-fontsize": {
"default": 10,
"description": "Controls the axis font size (labels and values)",
"type": "uint16"
},
"bar-norm": {
"default": true,
"description": "If set to true will normalize the distribution between 0-1",
"type": "bool"
},
...
},
"uses": {
"baselib": "plot/baselib/1"
}
}
Here are the description for each of the fields in the example above:
dataformat is the specific dataformat in the input of a plotter.
description gives a small description of the plotter.
language is the language in which the source code is implemented.
parameters are the given parameters to the source code to produce the figure. These parameters can be manipulated by the Plotterparameters. Default values for some of the parameters can be defined in a separate text file in the same directory as the JSON file. This text file has key/value format similar to a JSON file.
uses are the libraries used by the plotter.
The graphical interface of BEAT provides user-friendly editors to configure the JSON declaration definition of plotters when using BEAT locally. At the moment, defining new plotters on the platform is possible only under administrative privilege.
Plotter Types and Dataformats¶
At the moment there are three different types of plotters provided by BEAT platform; bar plots, ROC/DET plots, and scatter plots. However users can make their own plotters when using BEAT locally.
Each plotter accepts only a specific type of dataformat in their input. For example a bar plot only accepts plot/bar/1
dataformat. Here is the JSON declaration for this dataformat.
{
"#description": "Array of bar plots",
"data": [
0,
{
"label": "string",
"x": [
0,
"float64"
],
"y": [
0,
"float64"
]
}
]
}
and the scatter plot only accepts plot/scatter/1
dataformat which has the following JSON declaration:
{
"#description": "A set of 2-D scatter plots with X/Y and a label",
"data": [
0,
{
"label": "string",
"x": [
0,
"float64"
],
"y": [
0,
"float64"
]
}
]
}