By building machine learning directly into the mathematical transform, we can get much better results than simply apply machine learning to the radar, sonar, or lidar images themselves! Hence the adaptive chirplet transform as an important step toward achieving Simon Haykin's goal of radar, sonar, or lidar systems capable of cognition [1].
Background reading:
(Picture caption: Chirplet Transform and HuMachine Learning in VR/AR environment with the
SYSU x MannLab Lock-In Amplifier)