Twin Neural Network Regression is a Semi-Supervised Regression Algorithm

TL;DR

This paper introduces Twin Neural Network Regression (TNNR), a semi-supervised regression method that predicts differences between data points, enabling effective training with limited labeled data and achieving state-of-the-art results.

Contribution

The paper presents a novel semi-supervised regression algorithm, TNNR, which leverages difference predictions and loop constraints to improve accuracy with minimal labeled data.

Findings

Semi-supervised training significantly improves TNNR performance.

TNNR achieves state-of-the-art results in regression tasks.

Loop constraints help incorporate unlabeled data effectively.

Abstract

Twin neural network regression (TNNR) is a semi-supervised regression algorithm, it can be trained on unlabelled data points as long as other, labelled anchor data points, are present. TNNR is trained to predict differences between the target values of two different data points rather than the targets themselves. By ensembling predicted differences between the targets of an unseen data point and all training data points, it is possible to obtain a very accurate prediction for the original regression problem. Since any loop of predicted differences should sum to zero, loops can be supplied to the training data, even if the data points themselves within loops are unlabelled. Semi-supervised training improves TNNR performance, which is already state of the art, significantly.