search

automl / RoBO

RoBO: a Robust Bayesian Optimization framework
BSD 3-Clause "New" or "Revised" License
480 stars 134 forks source link

Tutorial for MTBO is not available #94

Open AnilRamachandran opened 6 years ago

AnilRamachandran commented 6 years ago

Can anyone please suggest, how objective function is defined for mtbo.

aaronkl commented 6 years ago

the interface looks the same as for Fabolas just that the additional variable is an integer defining the task rather than a continuous variable:

def objective(x, task):
    return function_value, cost

Also have a look here: https://github.com/automl/RoBO/blob/master/experiments/fabolas/run_mtbo.py

AnilRamachandran commented 6 years ago

Ok. thank you so much!

AnilRamachandran commented 6 years ago

I think the code for MTBO is for two task problem as in the experiments in MTBO paper. Is there any way I can run that code for multiple tasks? I mean how should I take the observations from 1,...(t-1) tasks when optimizing t_th task?