Abstraction and Model Simplification to Identify Interesting Data (RAMS)

Remote sensing platforms are often able to transfer only a small portion of all collected data to end-users, requiring significant manual effort to select the most relevant information for analysis.nbsp; To address this challenge, the ATC-NY team will develop Response Abstraction and Model Simplification (RAMS), a decision-support tool that assists scientists and automates remote and deep-space data collection for known events.nbsp; RAMS operates efficiently on remote sensing platforms by quantizing samples of telemetry data to enable highly parallel processing of Quantized Neural Network (QNN) operations.nbsp; RAMS also applies transfer learning and active learning techniques to train effective event detection models that reproduce human data-selection processes using a limited number of examples.nbsp; Using RAMS, scientists supporting the Magnetospheric Multiscale (MMS) mission identify several examples of target signals for magnetic reconnection events near the Earthrsquo;s magnetopause and magnetotail, which RAMS uses to automatically select such events in future data.