# Iterative Self-Learning: Semi-Supervised Improvement to Dataset Volumes and Model Accuracy

## TL;DR

This paper introduces an iterative semi-supervised learning method that leverages unlabelled data, thresholding, and ensemble decision systems to improve model accuracy and dataset volume, validated on standard and challenging image datasets.

## Contribution

It presents a new semi-supervised learning approach combining iterative cycles, learned thresholds, and ensemble methods for enhanced dataset utilization and model performance.

## Key findings

- Achieved state-of-the-art results on benchmark datasets
- Increased training data volume with unlabelled data improves accuracy
- Effective on both standard and complex image classification datasets

## Abstract

A novel semi-supervised learning technique is introduced based on a simple iterative learning cycle together with learned thresholding techniques and an ensemble decision support system. State-of-the-art model performance and increased training data volume are demonstrated, through the use of unlabelled data when training deeply learned classification models. Evaluation of the proposed approach is performed on commonly used datasets when evaluating semi-supervised learning techniques as well as a number of more challenging image classification datasets (CIFAR-100 and a 200 class subset of ImageNet).