Forrester¶
Forrester Task from the Profet paper.
This Forrester class is based on a synthetic function, whereas the ForresterTask is baseed on a meta-model trained on multiple such functions.
Klein, Aaron, Zhenwen Dai, Frank Hutter, Neil Lawrence, and Javier Gonzalez. “Meta-surrogate benchmarking for hyperparameter optimization.” Advances in Neural Information Processing Systems 32 (2019): 6270-6280.
- class orion.benchmark.task.forrester.Forrester(max_trials: int, alpha: float = 0.5, beta: float = 0.5)[source]¶
Task based on the Forrester function, as described in https://arxiv.org/abs/1905.12982
\[f(x) = ((lpha x - 2)^2) sin(eta x - 4)\]- Parameters
- max_trialsint
Maximum number of trials for this task.
- alphafloat, optional
Alpha parameter used in the above equation, by default 0.5
- betafloat, optional
Beta parameter used in the above equation, by default 0.5
Methods
call
(x)Define the black box function to optimize, the function will expect hyper-parameters to search and return objective values of trial with the hyper-parameters.
Return the search space for the task objective function
- call(x: float) List[Dict] [source]¶
Define the black box function to optimize, the function will expect hyper-parameters to search and return objective values of trial with the hyper-parameters.
This method should be overridden by subclasses. It should receive the hyper-parameters as keyword arguments, with argument names matching the keys of the dictionary returned by
get_search_space
.