Multi-label Ranking: Mining Multi-label and Label Ranking Data

TL;DR

Developments in the last demi-decade are surveyed, with a special focus on state-of-the-art methods in deep learning multi-label mining, extreme multi- label classification and label ranking.

Abstract

We survey multi-label ranking tasks, specifically multi-label classification and label ranking classification. We highlight the unique challenges, and re-categorize the methods, as they no longer fit into the traditional categories of transformation and adaptation. We survey developments in the last demi-decade, with a special focus on state-of-the-art methods in deep learning multi-label mining, extreme multi-label classification and label ranking. We conclude by offering a few future research directions.

Figures (4)

• Figure 1: High dimensionality in the output space. The number of instances (y-axis) with a given labelset (x-axis) in the Genebase dataset.
• Figure 2: Label correlation: a chord diagram gu2014circlize showing the concurrence of 13 labels in the Genebase dataset. The arcs represent label concurrence For example, the protein to the right of 12 o'clock (PDO-0196) co-occurs with only one other protein (PDO-0199) and that happens less frequently than other concurrences.
• Figure 3: Label imbalance. The number of instances (y-axis) with a given label (x-axis) in the Genebase dataset.
• Figure 4: Labelset size imbalance. The number of instances (y-axis) with a given number of labels (x-axis) in the Genebase dataset.

Theorems & Definitions (3)

• definition 1
• definition 2
• definition 3