Constructing Multiple Tasks for Augmentation: Improving Neural Image Classification With K-means Features

TL;DR

This paper introduces a novel multi-task learning method that constructs auxiliary tasks via unsupervised clustering and employs meta-learning to improve neural image classification, significantly outperforming existing methods.

Contribution

The paper proposes a meta-learning-based approach to effectively utilize unsupervised clustering for auxiliary task construction in multi-task learning.

Findings

Achieved over 9% improvement on Omniglot dataset.

Outperformed existing single-task and semi-supervised methods.

Demonstrated effectiveness across five image datasets.

Abstract

Multi-task learning (MTL) has received considerable attention, and numerous deep learning applications benefit from MTL with multiple objectives. However, constructing multiple related tasks is difficult, and sometimes only a single task is available for training in a dataset. To tackle this problem, we explored the idea of using unsupervised clustering to construct a variety of auxiliary tasks from unlabeled data or existing labeled data. We found that some of these newly constructed tasks could exhibit semantic meanings corresponding to certain human-specific attributes, but some were non-ideal. In order to effectively reduce the impact of non-ideal auxiliary tasks on the main task, we further proposed a novel meta-learning-based multi-task learning approach, which trained the shared hidden layers on auxiliary tasks, while the meta-optimization objective was to minimize the loss on the main task, ensuring that the optimizing direction led to an improvement on the main task. Experimental results across five image datasets demonstrated that the proposed method significantly outperformed existing single task learning, semi-supervised learning, and some data augmentation methods, including an improvement of more than 9% on the Omniglot dataset.