id	author	title	date	pages	extension	mime	words	sentences	flesch	summary	cache	txt
cord-127025-9ubhd4vf	Abraham, Louis	Crackovid: Optimizing Group Testing	2020-05-13		.txt	text/plain	4538	296	64	Our mathematical objective is designed such that the mixture tests it proposes to run in the lab will maximize the amount of information we gain on the ground truth once their lab results are revealed − in expectation, over the randomness of both imperfect tests and prior probabilities of infection per individual. This leads us to the following question: given an initial prior probability distribution p S over the secret, how should we select pool designs to test in the lab? Given numbers n & m, test characteristics tpr & tnr as well as prior probabilities of sample infection p i , the best multiset D of m pool designs is the one maximizing some score, like I(S, T (S, D)) or Confidence(S, T (S, D)). 9 Those prior probabilities can then be readily used by our approach to optimize the pool designs, and the ML system can gradually be improved as we gather more test results.	./cache/cord-127025-9ubhd4vf.txt	./txt/cord-127025-9ubhd4vf.txt