Semi-Supervised Learning, Causality and the Conditional Cluster Assumption
This work addresses a theoretical limitation in SSL for mixed causal settings, offering a potential algorithmic extension, but it appears incremental as it builds directly on prior causal insights without demonstrating empirical results.
The paper tackles the problem of extending semi-supervised learning (SSL) to scenarios involving both cause and effect features, such as predicting disease from risk factors and symptoms, by proposing to use information in the conditional distribution of effect features given causal features instead of the marginal distribution of all inputs.
While the success of semi-supervised learning (SSL) is still not fully understood, Schölkopf et al. (2012) have established a link to the principle of independent causal mechanisms. They conclude that SSL should be impossible when predicting a target variable from its causes, but possible when predicting it from its effects. Since both these cases are somewhat restrictive, we extend their work by considering classification using cause and effect features at the same time, such as predicting disease from both risk factors and symptoms. While standard SSL exploits information contained in the marginal distribution of all inputs (to improve the estimate of the conditional distribution of the target given inputs), we argue that in our more general setting we should use information in the conditional distribution of effect features given causal features. We explore how this insight generalises the previous understanding, and how it relates to and can be exploited algorithmically for SSL.