Possibly not usefull for DeepLearning enthousiasts, but the article by John Kay (economist) gives an example about simple versus complex modeling during the initial HIV epidemic in the eighties. More specifically he notes how a relatively simple model by two mathematicians was more succesfull than a complex model by the WHO (which took into account the latest country-by-country demographic data).
Maybe some food for though when considering parameters during the current pandemic 
https://www.johnkay.com/2020/03/15/coronavirus-a-black-swan/#_ftnref3