PyFlux 0.1.3 Release

posted in: Uncategorized | 1

Version 0.1.3 of PyFlux has been released on PyPi, including functionality for:

Models

  • ARIMA models – see here for a demonstration

Inference

  • Maximum Likelihood – see here for a demonstration
  • Maximum a Posteriori – see here for a demonstration
  • Laplace Approximation – see here for a demonstration
  • Metropolis-Hastings – see here for a demonstration

This is a “bare bones” package as of now. In the coming months, future releases will build on this basic foundation. The main priority right now is to expand model and inference options. This may mean optimization of existing features may lag behind a little, but we’ll get there!

Enjoy,

Ross

One Response

  1. Update to this:

    One of the critical issues with the current features is the inverse Hessian estimates being produced from MLE and MAP fitting methods. Replacing numerical BFGS with an analytical method for ARIMA and GAS models is a priority and will be updated in one of the next few versions.

Leave a Reply