Semi-Supervised Learning with Declaratively Specified Entropy Constraints
This work addresses the challenge of designing and combining semi-supervised learning strategies for researchers and practitioners, though it appears incremental as it builds on existing heuristics like co-training.
The authors tackled the problem of specifying strategies for semi-supervised learning by proposing a declarative method to model ensembles, constraints, and heuristics, achieving consistent improvements on benchmarks and a new state-of-the-art result on a difficult relation extraction task.
We propose a technique for declaratively specifying strategies for semi-supervised learning (SSL). The proposed method can be used to specify ensembles of semi-supervised learning, as well as agreement constraints and entropic regularization constraints between these learners, and can be used to model both well-known heuristics such as co-training and novel domain-specific heuristics. In addition to representing individual SSL heuristics, we show that multiple heuristics can also be automatically combined using Bayesian optimization methods. We show consistent improvements on a suite of well-studied SSL benchmarks, including a new state-of-the-art result on a difficult relation extraction task.