By now we have all heard of the challenges of modeling coronavirus spread. What we know today may need to be revised once more data becomes available (per Drs Birx and Fauci of CDC). Does that ring a bell in machine learning?

Perspective - if we draw a parallel between the R-naught* of covid-19 model and the b-factor in Arp’s hyperbolic decline curve for a well (and there exist many models to characterize well behavior), maybe we’d better appreciate the challenges CDC and virologists face.

B-story - there is not a single number, but a range of b’s (varying spatially and with basin as the variance in well productivity). Be the b’s gotten from science-based ML or analytical model (using multi-period modified Arp’s equations), learning with machine beats machine learning alone.

Hence the more pertinent question for covid-19 may not be which model is right, but how we make sense of R-naught. So if we can transfer model experience in “flattening” the decline curve in production, we may contribute to the understanding of virus spread and better manage the risk of living with coronavirus.

Any b-story to share?


*link to R-naught article.
https://www.scientificame...w-one-little-number/

Serendipity / perfect timing - https://www.pbs.org/newsh...s-and-what-they-cant