Frequently Asked Questions

Q: My computation terminates before sampling sufficient number of particles. How do I make it run longer?

A: All parameter inference functions accept an optional parameter max_iter, which specifies the number of iterations to run. See ABC Reference for more details.

Q: When running the example on my machine, emulation provides no performance benefit or even takes longer than emulation. How do I speed up emulation?

A: The best way to reduce the time it takes to run the emulation is to decrease the batch_size parameter. Please refer to docs for EmulatedABCRejection end EmulatedABCSMC. Machines with larger amount of RAM would be able to handle larger batches without performance degradation.

Q: How do I generate a plot that is similar to the one in the paper?

A: You will need to run ABC-SMC parameter estimation using both emulation and simulation for the same problem. Then, assuming you have Plots package installed, just run

julia> plot(emu_out, sim_out, true_params)

Use pyplot() backend for best results.

Q: How do I run the example notebooks locally?

A:

• Make sure that Jupyter is installed, along with Julia and its Jupyter Kernel.
• Clone or download GpABC.jl to your machine.
• Run jupyter from the examples directory of the local copy of GpABC.jl:

$ cd GpABC.jl/examples
$ jupyter notebook

• The first line of the first cell, that contains something like

import Pkg; Pkg.activate(@__DIR__); Pkg.instantiate()
using ...

will download all the dependencies.

All notebooks were tested under Julia 1.2