Semi-supervised Deep Kernel Learning: Regression with Unlabeled Data by Minimizing Predictive Variance

TL;DR

This paper introduces SSDKL, a semi-supervised regression method that combines neural networks and Gaussian processes to effectively utilize unlabeled data, improving performance on real-world tasks.

Contribution

The paper proposes a novel semi-supervised deep kernel learning approach that minimizes predictive variance, integrating neural networks with Gaussian processes for regression.

Findings

Outperforms supervised deep kernel learning on various tasks

Leverages unlabeled data to improve regression accuracy

Demonstrates effectiveness over VAT and mean teacher methods

Abstract

Large amounts of labeled data are typically required to train deep learning models. For many real-world problems, however, acquiring additional data can be expensive or even impossible. We present semi-supervised deep kernel learning (SSDKL), a semi-supervised regression model based on minimizing predictive variance in the posterior regularization framework. SSDKL combines the hierarchical representation learning of neural networks with the probabilistic modeling capabilities of Gaussian processes. By leveraging unlabeled data, we show improvements on a diverse set of real-world regression tasks over supervised deep kernel learning and semi-supervised methods such as VAT and mean teacher adapted for regression.