Hyperactive
Hyperactive copied to clipboard
SimonBlanke
An optimization and data collection toolbox for convenient and fast prototyping of computationally expensive models.
Currently, we have adopted the `sktime` parallelization backend dispatcher, inheriting the issues with `ray` and testing `ray`, e.g., long runtimes, sporadic crashes in the CI - parallel issue in `sktime`...
Adds two benchmark functions (formerly used in the defunct `scikit-optimize` package): * Hartmann function * Branin-Hoo function
The `sktime` `ForecastingOptCV` provides basic tuning abilities, but is lacking some parameters that `sktime` native tuners have, e.g., `ForecastingGridSearchCV`. These should be added: * `tune_by_instance` and `tune_by_variable` parameters * attributes...
It would be nice to have callbacks for `BaseOptimizer` or `BaseExperiment`. One common use case is logging the sequence of parameters and parameter values obtained throughout the optimization. My current...
The test surfaces in `surfaces` have an argument `sleep` that introduces latency, e.g., for testing purposes. https://github.com/SimonBlanke/Surfaces I think this is a neat idea, in `hyperactive` I would deal with...
I think we will need tags for optimizers that tell the user - and interacting parts of the framework - what types it can accept, e.g., integer/real numbers, or string/categorical....
A first composition for experiments, `Negative` - this flips the signs in an experiment. This should also be considered as a proof-of-concept for how composition of experiments would look like....
`test_trafo_1` fails sporadically with a timings check. Example traceback below. If this is not a strict requirement, we should perhaps remove this or make it more lenient? ``` ================================== FAILURES...
Metadata
Owner
Metadata
An optimization and data collection toolbox for convenient and fast prototyping of computationally expensive models.