One vs Previous and Similar Classes Learning -- A Comparative Study

TL;DR

This paper compares three new paradigms for updating multi-class classifiers efficiently without retraining from scratch, demonstrating faster update times and comparable accuracy, especially on larger datasets.

Contribution

It introduces three novel learning paradigms for incremental multi-class classification updates, reducing retraining time while maintaining performance.

Findings

Proposed paradigms are faster at updating than the baseline.

Two paradigms are also faster at training from scratch on large datasets.

Classification performance remains comparable to the baseline.

Abstract

When dealing with multi-class classification problems, it is common practice to build a model consisting of a series of binary classifiers using a learning paradigm which dictates how the classifiers are built and combined to discriminate between the individual classes. As new data enters the system and the model needs updating, these models would often need to be retrained from scratch. This work proposes three learning paradigms which allow trained models to be updated without the need of retraining from scratch. A comparative analysis is performed to evaluate them against a baseline. Results show that the proposed paradigms are faster than the baseline at updating, with two of them being faster at training from scratch as well, especially on larger datasets, while retaining a comparable classification performance.