Author: Marcus Ludwig; Louis-Félix Nothias; Kai Dührkop; Irina Koester; Markus Fleischauer; Martin A. Hoffmann; Daniel Petras; Fernando Vargas; Mustafa Morsy; Lihini Aluwihare; Pieter C. Dorrestein; Sebastian Böcker
Title: ZODIAC: database-independent molecular formula annotation using Gibbs sampling reveals unknown small molecules Document date: 2019_11_16
ID: 03uonbrv_107
Snippet: Burn-in and number of Gibbs sampling epochs. We determined a reasonable number of Gibbs sampling iterations using the dendroides dataset. One iteration, also called epoch, is dened as one round in which each compound is updated once by choosing a new active molecular formula candidate. We run 10 independent Markov chains, see The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/842740 doi: bioRxiv .....
Document: Burn-in and number of Gibbs sampling epochs. We determined a reasonable number of Gibbs sampling iterations using the dendroides dataset. One iteration, also called epoch, is dened as one round in which each compound is updated once by choosing a new active molecular formula candidate. We run 10 independent Markov chains, see The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/842740 doi: bioRxiv preprint correct annotations at a specic epoch increases quickly for most Markov chains until the chain seems to stay in a local optimum. We note that this number of correct molecular formula is determined at each epoch whereas ZODIAC scores are computed from the average over many epochs. From this data, we estimated a burn-in of 1,000 epochs and sampling of 2,000 iterations. Larger values increase running times but should never worsen results.
Search related documents:
Co phrase search for related documents, hyperlinks ordered by date