PP: Imaging time-series to improve classification and imputation

From: University of Maryland

encode time series as different types of images.

reformulate features of time series as visual clues.

three representations for encoding time series as images: Gramian angular summation fields/ Gramian angular difference fields and Markov transition fields.

Recently, researchers are trying to build different network structures from time series for visual inspection or designing distance measures.

build a weighted adjacency matrix is extracting transition dynamics from the first order Markov matrix.

time series ---------> topological properties; but it remains unclear how these topological properties relate to the original time series since they have no exact inverse operations.

time series ----> images ----> tailed CNN for classification

Conclusion:

We aim to further apply our time series models in real world regression/imputation and anomaly detection tasks.