stad

Dimensionality reduction through Simplified Topological Abstraction of Data


Keywords
dimensionality-reduction, visualisation, visualization
License
MIT
Install
pip install stad==2.0.1

Documentation

pySTAD - Python implementation of Simplified Topological Abstraction of Data

Installation

pip install stad

Usage

The input to stad is a normalised distance matrix (i.e. with values between 0 and 1). Optionally, you can also provide an array of values for each datapoint that can be used in the lens.

Let's for example look at the five circles dataset that is used in the example script below. Without a lens, a stad analysis will reveal a circle with four spikes; with a lens each of these spikes itself also becomes a circle (as in the picture).

The data for this dataset looks like this:

x,y,hue
377,566,#1F988B
362,589,#21A585
350,607,#29AF7F
104,977,#20928C
124,978,#26818E
118,956,#1F9E89
...

Here's a complete script to create this graph:

import stad
import pandas as pd

## Load the data
url = 'https://gist.githubusercontent.com/jandot/a84c0505cdc8008a6e5ae5032532a39f/raw/d834527117fd204d33486998d10290251354d013/five_circles.csv'
data = pd.read_csv(url, header=0)

## Extract the values we want to use in our distance, the lens, and optional features
values = data[['x','y']].values.tolist()
lens = data['hue'].map(lambda x:stad.hex_to_hsv(x)[0]).values
xs = data['x'].values.tolist()
ys = data['y'].values.tolist()
hues = data['hue'].values.tolist()

## Create the distance matrix in the high_dimensional space. This can be using
## cosine distance, euclidean, or any other.
highD_dist_matrix = stad.calculate_highD_dist_matrix(values)

## Run STAD and show the result
g = stad.run_stad(highD_dist_matrix, lens=lens, features={'x':xs, 'y':ys, 'hue': hues})
stad.draw_stad(g)