BLIND DEINTERLEAVING OF SIGNALS IN TIME SERIES WITH SELF-ATTENTION BASED SOFT MIN-COST FLOW LEARNING
IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), ELECTR NETWORK, 6 - 11 June 2021, pp.3295-3299, (Full Text)
- Publication Type: Conference Paper / Full Text
- Doi Number: 10.1109/icassp39728.2021.9415025
- Country: ELECTR NETWORK
- Page Numbers: pp.3295-3299
- Keywords: Deinterleaving, attention, min-cost flow, IMPROVED ALGORITHM
- Middle East Technical University Affiliated: Yes
Abstract
We propose an end-to-end learning approach to address deinterleaving of patterns in time series, in particular, radar signals. We link signal clustering problem to min-cost flow as an equivalent problem once the proper costs exist. We formulate a bi-level optimization problem involving min-cost flow as a sub-problem to learn such costs from the supervised training data. We then approximate the lower level optimization problem by self-attention based neural networks and provide a trainable framework that clusters the patterns in the input as the distinct flows. We evaluate our method with extensive experiments on a large dataset with several challenging scenarios to show the efficiency.