Multi-fidelity refers to running an algorithm on multiple budgets (such as number of epochs or
subsets of data) and thereby evaluating the performance prematurely. You can run a multi-fidelity optimization
when using Successive Halving or
Hyperband. Hyperband is the default intensifier in the
multi-fidelity facade and requires the arguments
max_budget in the scenario if no instances are used.
In general, multi-fidelity works for both real-valued and instance budgets. In the real-valued case,
the budget is directly passed to the target function. In the instance case, the budget is not passed to the
target function but
max_budget are used internally to determine the number of instances of
each stage. That’s also the reason why
max_budget are not required when using instances:
max_budget is simply the max number of instances, whereas the
min_budget is simply 1.
Please have a look into our multi-fidelity examples to see how to use multi-fidelity optimization in real-world applications.