This is the code for constructing the Thresholding After Random Projection (TARP) benchmark for a detection problem using labeled data. The benchmark is a curve in a two-dimensional plane whose x-axis is the Area Above the ROC Curve (AAC) and whose y-axis is (either training or testing) computational cost (CC). We build the curve by first estimating the AAC and CC for a sequence of TARP-based detection methods with increasing complexity. Each TARP-based detection method defines a point in the plane given by the AAC and the CC of that method. The curve is obtained by interpolating the sequence of points corresponding to the sequence of TARP method performance (AAC-CC) with a straight line. The code herein produces a CSV file containing the sequence of points use to build this benchmark curves.
Cite this work
Researchers should cite this work as follows: