Semi-Supervised Learning: the Case When Unlabeled Data is Equally Useful
This provides theoretical insights for researchers in machine learning, showing conditions where unlabeled data enhances learning rates, but it is incremental as it builds on existing semi-supervised frameworks.
The paper tackles the problem of semi-supervised learning by analyzing statistical models with continuous parameters, showing that under certain distribution conditions, unlabeled data can be as useful as labeled data in terms of learning rate, with rates scaling as O(1/n) when m∼n and O(1/n^{1+γ}) when m∼n^{1+γ} for γ>0, compared to supervised learning's O(1/n).
Semi-supervised learning algorithms attempt to take advantage of relatively inexpensive unlabeled data to improve learning performance. In this work, we consider statistical models where the data distributions can be characterized by continuous parameters. We show that under certain conditions on the distribution, unlabeled data is equally useful as labeled date in terms of learning rate. Specifically, let $n, m$ be the number of labeled and unlabeled data, respectively. It is shown that the learning rate of semi-supervised learning scales as $O(1/n)$ if $m\sim n$, and scales as $O(1/n^{1+γ})$ if $m\sim n^{1+γ}$ for some $γ>0$, whereas the learning rate of supervised learning scales as $O(1/n)$.