How to regenerate the equations from hof files ?

[qizzhao]

Hi,
I am using SymbolicRegression package in julia, which will generate and store a hof file after training. If I restart the julia, how could I regeneration those equations from hof file? I want to use them to do predictions on a different dataset.

I know we could use model.__from_file__ in python. But please anyone tell me how to do in Julia.

Thank you

[gm89uk]

using SymbolicRegression, CSV, MLJ
guesses = CSV.read("path\\hall_of_fame.csv", DataFrame).Equation
model = SRRegressor(
   #...
    guesses = guesses,
    fraction_replaced_guesses=0.0, #unless you want the search to keep referring back to your guesses 
)

mach = machine(model, X, y)
fit!(mach)

This will however, re-evolve the equations to match your new dataset, presumably this is what you want, rather than a post-hoc test.

[qizzhao]

Thank you so much!
I have approached your suggestion and it is working!
And there are 2 related questions:

1. How to make sure the output of fit!(mach) will return all equations of hof.csv? It seems it will ignore part of hof's equations after fitting. (I tried to set ntop=100, but not working)
2. Is there any way that I could speed up the startup time of fit!(mach)? It took a long waiting time. (I tried to set timeout_in_second=1, but not working)

[MilesCranmer]

@qizzhao regarding 1, it should be report(mach) after running fit!. If it doesn't show all the equations, could you make a bug report with code to reproduce?

2. I think this might also be worth submitting a bug report! Unless you mean the very first call to fit!, in which case, the solution is to keep Julia open between calls, so that the JIT compilation doesn't need to run repeatedly