Semi-Supervised Learning

Synonyms

Co-training; Learning from labeled and unlabeled data; Transductive learning

Definition

Semi-supervised learning uses both labeled and unlabeled data to perform an otherwise supervised learning or unsupervised learning task.

In the former case, there is a distinction between inductive semi-supervised learning and transductive learning. In inductive semi-supervised learning, the learner has both labeled training data $\{(\mathbf{x}_i, y_i)\}_{i=1}^l {\mathop{\sim}^{iid}} p(\mathbf{x},y)$ and unlabeled training data $\{{\mathbf{x}_{i}\}}_{i\,=\,l+1}^{l+u} {\mathop{\sim}^{iid}}p(\mathbf{x})$, and learns a predictor $f : \mathcal{X}\mapsto \mathcal{Y}$, $f \in \mathcal{F}$ where $\mathcal{F}$ is the hypothesis space. Here $\mathbf{x} \in \mathcal{X}$ is an input instance, $y \in \mathcal{Y}$ its target label (discrete for classification or continuous for regression), p(x, y) the unknown joint distribution and p(x) its marginal, and typically l ≪ u. The goal is to...