prosphera 1.0.3

Visualize Multidimensional Data on a Sphere

This tool utilises sophisticated PCA with a cosine kernel to generate informative visualisations of multi-dimensional data in three-dimensional space. Following the PCA process, the data is normalised by shifting each point to a centroid and making it the unit norm. To enhance the visualisation, vectors are additionally scaled with precision to move the farthest points closer to the surface of the sphere. The outcome is an engaging and instinctive representation of the data in spherical format. The tool initiates interactive visualisations in a new tab of your default web browser, facilitating data exploration and analysis.

Basic Usage

Init visualizer

import pandas as pd
from sklearn import datasets
from sklearn.datasets import make_blobs
from prosphera.projector import Projector

# Instantiate the class
visualizer = Projector()

Generated dataset

# Generate data
data, labels = make_blobs(
    n_samples=5000,
    centers=50,
    n_features=25,
    random_state=1234)

# Call the visualize method to generate and display the visualization
visualizer.project(
    data=data,
    labels=labels)

Browser tab:

Wine dataset

wine = datasets.load_wine()

visualizer.project(
    data=wine['data'],
    labels=wine['target'])

Browser tab:

Cancer dataset

cancer = datasets.load_breast_cancer()

visualizer.project(
    data=cancer['data'],
    labels=cancer['target'])

Browser tab:

Digits dataset (no labels)

digits = datasets.load_digits(n_class=5)

visualizer.project(
    data=digits['data'],
    meta=digits['target'])

Browser tab:

Digits dataset (apply labels)

visualizer.project(
    data=digits['data'],
    labels=digits['target'])

Browser tab:

Housing dataset (labels from 'age')

housing = datasets.fetch_california_housing()

visualizer.project(
    data=housing['data'],
    labels=pd.qcut(housing['data'][:, 1], 5).astype(str))

Browser tab:

Change renderer

You can set renderer as visualizer = Projector(renderer='iframe') to save the plot locally as HTML. Available renderers:

• 'jupyterlab'
• 'vscode'
• 'notebook'
• 'kaggle'
• 'colab' and others