Dynimp: Dynamic Imputation for Wearable Sensing Data through Sensory and Temporal Relatedness

Abstract

In wearable sensing applications, data is inevitable to be irregularly sampled or partially missing, which pose challenges for any downstream application. An unique aspect of wearable data is that it is time-series data and each channel can be correlated to another one, such as x, y, z axis of accelerometer. We argue that traditional methods have rarely made use of both times-series dynamics of the data as well as the relatedness of the features from different sensors. We propose a model, termed as DynImp, to handle different time point's missingness with nearest neighbors along feature axis and then feeding the data into a LSTM-based denoising autoencoder which can reconstruct missingness along the time axis. We experiment the model on the extreme missingness scenario (> 50% missing rate) which has not been widely tested in wearable data. Our experiments on activity recognition show that the method can exploit the multi-modality features from related sensors and also learn from history time-series dynamics to reconstruct the data under extreme missingness.

Multi-Branching Temporal Convolutional Network With Tensor Data Completion for Diabetic Retinopathy Prediction

“…The core-TCN can be treated as a feature extractor or backbone of the network, and the MB layer is the output classifier. Existing literature has shown that the classifier component of the network suffers more from the imbalanced data issue, while the backbone portion is less susceptible during training [54], [55]. As such, we propose to train the core-TCN with the whole dataset and use a balanced sub-dataset to train each branching output in the MB layer to alleviate the impact of imbalanced data on the classifier.…”