If you have matlab's Parallel computing toolbox installed, you can open or close a parallel pool from the file menu. YADA always makes use of your default profile for the pool.
When you have opened a parallel pool, YADA makes use of the parfor function when running certain loops provided that it can use at least 2 computational threads (number of workers). The tools currently supported for this are listed below:
• | The mutation step of the SMC with likelihood or data tempering posterior sampler for DSGE and DSGE-VAR models, as well as the correction step of the SMC with data tempering posterior sampler. |
• | The computation of importance weights and t-densities when using importance sampling based on the MitISEM algorithm. |
• | CRPS and ES computations for DSGE models under rational expectations and subject to adaptive learning. |
• | Unconditional predictive distributions for DSGE models under rational expectations and subject to adaptive learning. |
• | Conditional predictive distributions for DSGE models under rational expectations. |
• | Zero (effective) lower bound computations. |
• | Inverse Hessian with finite differences. |
• | Predictive likelihood estimation and marginal predictive moments estimation with prior and posterior draws for DSGE-VAR models, and for DSGE models under rational expectations and adaptive learning. |
Page url:
https://www.texlips.net/yada/html/index.html?parallel_computing_toolbox.htm