Title: Kalman filter: A Julia implementation
Authors: Michel Juillard - Bank of France (France) [presenting]
Abstract: The Kalman filter and smoother is central to the estimation of linear state space models and of linearized DSGE models in particular. Estimation techniques, based either on optimization or on MCMC, require a large number of evaluations of the Kalman filer. This is the most time consuming single step in practical DSGE modeling. It is therefore essential to have a fast implementation of this algorithm. The Julia language is known for its speed and provides various tools to optimize code. It is a natural testing bench. Two variants of the filter are also explored: the Chandrasekhar recursion and the diffuse filter. An alternative implementation with PaddedMatrices is also presented.