🤖 AI Summary
This paper addresses missing value imputation and forecasting for irregular multivariate time series—data lacking regular sampling intervals, temporal alignment, or structural stationarity. Method: We propose AGG, a cycle-free graph attention network that makes no assumptions about temporal or spatial regularity. AGG jointly encodes measurements, timestamps, and channel-specific features into node embeddings; introduces a conditional attention mechanism to dynamically synthesize nodes; and constructs a data-driven homogeneous relation graph by modeling asynchronous inter-variable dependencies. Contribution/Results: AGG is the first to employ conditional attention for dynamic node synthesis and to frame asynchronous temporal dependency modeling as a data augmentation strategy. Evaluated on Beijing Air Quality, PhysioNet ICU-2012, and UCI Localisation benchmarks, AGG achieves state-of-the-art performance across imputation, classification, and forecasting tasks—significantly outperforming existing attention-based models.
📝 Abstract
We introduce the asynchronous graph generator (AGG), a novel graph attention network for imputation and prediction of multi-channel time series. Free from recurrent components or assumptions about temporal/spatial regularity, AGG encodes measurements, timestamps and channel-specific features directly in the nodes via learnable embeddings. Through an attention mechanism, these embeddings allow for discovering expressive relationships among the variables of interest in the form of a homogeneous graph. Once trained, AGG performs imputation by emph{conditional attention generation}, i.e., by creating a new node conditioned on given timestamps and channel specification. The proposed AGG is compared to related methods in the literature and its performance is analysed from a data augmentation perspective. Our experiments reveal that AGG achieved state-of-the-art results in time series imputation, classification and prediction for the benchmark datasets emph{Beijing Air Quality}, emph{PhysioNet ICU 2012} and emph{UCI localisation}, outperforming other recent attention-based networks.
