Welcome to the repo that will hopefully facilitate progress and discussion in the application of probabilistic programming to transport and traffic modelling problems.
Transport Modelling is a discipline that uses forecasting and prediction on imperfect models extensively and liberally. Despite this, I still believe in it, because some information is better than none. Nonetheless I feel the industry is due for an introduction to Bayesian methods particularly all the new advancements in methods and useability. I see it having significant benefits for the transport modelling field.
Initial phases of this library will heavily leverage methods established in the text book - Statistical Rethinking (Second Edition) and the port of the source code to Numpyro, and some other relevant papers. Essentially the repo will contain toy example implimentations of transport modelling problems solve using bayes stats in jupyter notebooks. The plan then is to progressively build up a library of helper classes/functions for methods or techniques that prove usefull.
Anticipated benefits of bayesian methods includes: supporting structured human reasoning, python syntax, natural uncertainty quantification, pooling / shrikage for data with varying sample sizes over index variables etc
This package is open source (GPL3).
The content of this package does not contain or represent the views of any organisation.