Geoscience Reference
In-Depth Information
84 CHAPTER 8. THEORY ANDOUTLOOK
What is next for semi-supervised learning? We wish the following questions will one day be
solved: How to efficiently incorporate massive, even unlimited, amounts of unlabeled data? How
to ensure that semi-supervised learning always outperforms supervised learning, by automatically
selecting the best semi-supervised model assumption and model parameters? How to design new
semi-supervised learning assumptions, perhaps with inspiration from cognitive science? Answering
these questions requires innovative research effort. We hope you will join the endeavor.
