2014/06/21 by Rushil Anirudh, Pavan Turaga, Anirudh, Rushil +1 · 1 citation
Biochemistry, Genetics and Molecular Biology · Engineering · Medicine · #Advanced biosensing and bioanalysis techniques #Biosensors and Analytical Detection #Computer Vision and Pattern Recognition (cs.CV) #Data-Driven Disease Surveillance #FOS: Computer and information sciences #SARS-CoV-2 detection and testing
paper · pdf · doi:10.48550/arxiv.1406.5653
openalex publication_date 2014/06/21 · openalex created_date 2022/02/24 · openalex updated_date 2026/07/28
In this paper, we study the problem of `test-driving' a detector, i.e.\nallowing a human user to get a quick sense of how well the detector generalizes\nto their specific requirement. To this end, we present the first system that\nestimates detector performance interactively without extensive ground truthing\nusing a human in the loop. We approach this as a problem of estimating\nproportions and show that it is possible to make accurate inferences on the\nproportion of classes or groups within a large data collection by observing\nonly 5-10 % of samples from the data. In estimating the false detections (for\nprecision), the samples are chosen carefully such that the overall\ncharacteristics of the data collection are preserved. Next, inspired by its use\nin estimating disease propagation we apply pooled testing approaches to\nestimate missed detections (for recall) from the dataset. The estimates thus\nobtained are close to the ones obtained using ground truth, thus reducing the\nneed for extensive labeling which is expensive and time consuming.\n