Figure 8.

A sample execution plan for Cortex around an example using only geolocation data (1) the clinician or researcher creates an aggregate operation; (2) Cortex transparently interposes the correct feature layers by creating a dependency graph of data and executes each atomic operation (ie, independent of external variables) in the order it computes to be most efficient; (3) any raw sensor data are transparently cached during execution; and (4) as multiple operations require the same raw sensor data, Cortex blocks their execution until the cached data becomes available, to avoid duplicate downloads, wasted computation, and oversaturation of network bandwidth.