Recurrent Halting Chain for Early Multi-label Classification

Abstract

Early multi-label classification of time series, the assignment of a label set to a time series before the series is entirely observed, is critical for time-sensitive domains such as healthcare. In such cases, waiting too long to classify can render predictions useless, regardless of their accuracy, while predicting prematurely can result in potentially costly erroneous results. When predicting multiple labels (for example, types of infections), dependencies between labels can be learned and leveraged to improve overall accuracy. Together, reliably predicting the correct label set of a time series while observing as few timesteps as possible is challenging because these goals are contradictory in that fewer timesteps often means worse accuracy. To achieve early yet sufficiently accurate predictions, correlations between labels must be accounted for since direct evidence of some labels may only appear late in the series. We design an effective solution to this open problem, the Recurrent Halting Chain (RHC), that for the first time integrates key innovations in both Early and Multi-label Classification into one multi-objective model. RHC uses a recurrent neural network to jointly model raw time series as well as correlations between labels, resulting in a novel order-free classifier chain that tackles this time-sensitive multi-label learning task. Further, RHC employs a reinforcement learning-based halting network to decide at each timestep which, if any, classes should be predicted, learning to build the label set over time. Using two real-world time-sensitive datasets and popular multi-label metrics, we show that RHC outperforms recent alternatives by predicting more-accurate label sets earlier. CCS CONCEPTS• Computing methodologies → Neural networks; Supervised learning by classification;

One Positive Label is Sufficient: Single-Positive Multi-Label Learning with Label Enhancement

“…In multi-label learning (MLL), each example is associated with a number of valid labels simultaneously. To cope with an output space which is exponential in size to the number of class labels, numerous approaches propose exploiting label correlations to improve the learning process [14,42,31]. The simplest one is the first-order type, which disassembles the MLL problem into a number of binary classification problems [2,43].…”

Early and Revocable Time Series Classification

“…(Rußwurm et al (2019)) proposed a trainable framework for early classification of time series that can be fine-tuned end-to-end using standard gradient back-propagation. (Martinez et al (2018)) and (Hartvigsen et al (2019)) approach early classification as a reinforcement learning problem, a perspective extended to multi-label classification in (Hartvigsen et al (2020)).…”

Robust Recurrent Classifier Chains for Multi-Label Learning with Missing Labels

“…RCCs have gained great popularity in the recent literature [20,43,54]. These methods have been applied to many multi-label tasks, including early classification [20], text mining [30], and computer vision [54], while numerous extensions of classifier chains have also been proposed [20,43,54]. However, no RCC model to date has been equipped to handle missing labels.…”