Changelog
5004c56 - Reimplement "Work in Progress" to measure the distance with 1-WD or EMD We are interested in penalizing not only the differences in the mass distribution but their distance, similarly to other metrics, because it's not the same to estimate the same WiP displaced 2h than 10h. (David Chapela de la Campa)
c11fef0 - Fix normalization of the measured distances When normalizing a measured distance, if the maximum observed value is 0 (all observations are in the first bin of the "histogram"), set the measured distance as normalized value instead of 0. Wasserstein will return 0 because if all observations are in the same bin in both "histograms", the PDFs are the same. EMD will return a difference if there's a difference in mass (desired). (David Chapela de la Campa)
3712a8d - Refactor Circadian Workforce to compute the average of active workers per hour and day of the week Realized that if we only aggregate the number of active workers, the same simulation scenario (w.r.t. the resource model) having a duration of 4 months vs 2 months (for example, because of different activity durations) would result in two very different workforces. Our aim is to compare the weekly workforce of one log with another, so by computing the average we mitigate this error. (David Chapela de la Campa)