informed machine learning

We introduce the paradigm of informed meta-learning, a novel approach to inductive bias specification based on human knowledge represented in any form, including unstructured natural language.