Picasso: Model-free Feature Visualization

TL;DR

Picasso introduces a novel model-free visualization method that displays thousands of features and their interactions in a single image, aiding in feature browsing and curation for large ML datasets.

Contribution

The paper presents a new visualization technique that efficiently represents large feature sets and their interactions without relying on model-based methods.

Findings

Visualizes thousands of features in one image

Conveys feature interactions through spatial positioning

Facilitates feature subset exploration

Abstract

Today, Machine Learning (ML) applications can have access to tens of thousands of features. With such feature sets, efficiently browsing and curating subsets of most relevant features is a challenge. In this paper, we present a novel approach to visualize up to several thousands of features in a single image. The image not only shows information on individual features, but also expresses feature interactions via the relative positioning of features.