ji

A toolbox for planar projections (2D embeddings) of high-dimensional data.

To install: pip install ji

Overview

The ji package provides a suite of tools for embedding high-dimensional data into two dimensions using various techniques, including PCA, Isomap, LLE, t-SNE, and more. It is useful for visualization, data compression, and as a preprocessing step for machine learning.

Main Features

• Dimensionality Reduction: Project high-dimensional data to 2D using a variety of methods.
• Visualization: Includes a function to create scatter plots of the embedded data.
• Extensible: Easily add more embedding methods or custom plotting functions.

Usage Examples

Demo: Listing and Using Planarizers

from ji import planarizers
# List available planarizers
print(sorted(planarizers))
# Prepare some data
X = [[1,2,3,4],[4,3,2,1],[0,0,0,0],[1,1,1,1],[2,2,2,2],[3,3,3,3],[4,4,4,4]]
y = [0,1,2,0,1,2,0]
# Use a planarizer that only uses X
pts = planarizers['identity'](X)
print(pts)
# Use a planarizer that arranges points in a circle
pts2 = planarizers['circular'](X)
print(pts2)

Using the analyze Function

The analyze function runs several planarizers and visualizes the results. It uses the planarizers registry, so you can easily add or use custom methods.

from ji import analyze
# Assuming X and y are already defined
analyze(X, y)

If you do not specify X and y, analyze will load a pre-defined digits dataset and perform the analysis on it:

from ji import analyze
analyze()

Custom Plot Function

You can pass a custom plotting function to analyze:

from ji import analyze
import matplotlib.pyplot as plt

def custom_plot(X, y):
    plt.scatter(X[:, 0], X[:, 1], c=y, cmap='viridis', edgecolor='k')
    plt.show()

analyze(plot_fun=custom_plot)

Testing

A test/demo function is provided:

from ji import test_planarizers
test_planarizers()

Documentation

scatter_plot(X, y)

Plots a simple scatter plot of the data.

• Parameters:
  • X: 2D numpy array, the data points to plot.
  • y: 1D numpy array, the target labels for coloring the points.

analyze(X=None, y=None, plot_fun=scatter_plot, data_name='data', methods=None)

Main function to perform and visualize various embeddings using the planarizers registry.

• Parameters:
  • X: 2D numpy array, the input data. If None, a default dataset (digits) is loaded.
  • y: 1D numpy array, the target labels corresponding to X. Required if X is not None.
  • plot_fun: Function used to plot the embedding. Defaults to scatter_plot.
  • data_name: String, a name for the dataset to be displayed in plot titles.
  • methods: List of planarizer names to use (defaults to a subset).

This package is designed to be a practical tool in both educational and research settings for those interested in machine learning, data science, and pattern recognition.