I show how the accuracy of production decline curve forecasting--a widely used but ill-posed inverse problem--can be improved by using a Bayesian formulation and machine learning.

Please contact author for more info: jbm@mit.edu