The likelihood annealing SMC approach for the linear regression example is currently based on using predetermined temperature schedules, MCMC kernels and number of MCMC repeats. These choices were all based on what worked well in previous runs, and in general this sort of information is not available.
One of the main benefits of using SMC in this context is that it is naturally adaptive. Although this example is fairly simple (purposely and for a few reasons), showing examples of adaptation will hopefully help to illustrate that the library if flexible enough for more complex problems.
The likelihood annealing SMC approach for the linear regression example is currently based on using predetermined temperature schedules, MCMC kernels and number of MCMC repeats. These choices were all based on what worked well in previous runs, and in general this sort of information is not available.
One of the main benefits of using SMC in this context is that it is naturally adaptive. Although this example is fairly simple (purposely and for a few reasons), showing examples of adaptation will hopefully help to illustrate that the library if flexible enough for more complex problems.