Abstract
A large number and diversity of techniques have been offered in theliterature in recent years for solving multi-label classification tasks,including classifier chains where predictions are cascaded to other models asadditional features. The idea of extending this chaining methodology tomulti-output regression has already been suggested and trialed: regressorchains. However, this has so-far been limited to greedy inference and hasprovided relatively poor results compared to individual models, and of limitedapplicability. In this paper we identify and discuss the main limitations,including an analysis of different base models, loss functions, explainability,and other desiderata of real-world applications. To overcome the identifiedlimitations we study and develop methods for regressor chains. In particular wepresent a sequential Monte Carlo scheme in the framework of a probabilisticregressor chain, and we show it can be effective, flexible and useful inseveral types of data. We place regressor chains in context in general terms ofmulti-output learning with continuous outputs, and in doing this shedadditional light on classifier chains.