skripts.VAE package

Submodules

skripts.VAE.VAE_interpret module

skripts.VAE.VAE_smac module

Hyperparameter tuning for a variational autoencoder for flow-injection analysis with SMAC.

class skripts.VAE.VAE_smac.FIA_VAE_tune(data, test_size: float, configuration_space: ConfigurationSpace, model_builder, log_dir: str, batch_size: int = 16, verbosity: int = 0, device: str = 'cpu', name: str = 'smac_vae')[source]

Bases: object

Class for running the SMAC3 tuning.

train(config: Configuration, seed: int = 0, budget: int = 25) → float[source]

Train the model for the number of defined steps.

Parameters:

  • config (Configuration) – Configuration to be trained upon.

  • seed (int, optional) – Initializing seed, defaults to 0

  • budget (int, optional) – Number of epochs to be used in training, defaults to 25

Returns:

Average loss of the model.

Return type:

float

skripts.VAE.VAE_smac.ask_tell_optimization(facade, smac_model, n: int = 10, verbosity: int = 0)[source]

Run the training run for n steps in a more verbose mode.

Parameters:

  • facade (smac.AbstractFacade) – Facade used by SMAC.

  • smac_model (Keras model) – SMAC model that is used

  • n (int, optional) – number of verbose runs, defaults to 10

  • verbosity (int, optional) – verbosity of output, defaults to 0

skripts.VAE.VAE_smac.main(args)[source]

Hyperparameter optimization with SMAC3

Parameters:

args (dictionary-like) – Arguments from shell. See VAE_smac.py –help for more information.

skripts.VAE.VAE_smac.run_optimization(facade, smac_model, verbose_steps: int = 10, verbosity: int = 0)[source]

Perform optimization run with SMAC facade.

Parameters:

  • facade (smac.AbstractFacade) – SMAC facade

  • smac_model (Keras model) – Model to supply for training

  • verbose_steps (int, optional) – number of steps to be returned in a more verbose fashion, defaults to 10

  • verbosity (int, optional) – Level of verbosity, defaults to 0

Returns:

Best hyperparameter combinations

Return type:

Union[list, dict, ConfigSpace.Configuration]

skripts.VAE.VAE_smac.save_runtime(run_dir, verbosity: int = 0)[source]

Saves the runtime Dataframe.

Parameters:

  • run_dir (str) – Directory for saving the run.

  • verbosity (int, optional) – Level of verbosity, defaults to 0

Returns:

Runtimes

Return type:

pandas.DataFrame

skripts.VAE.VAE_smac.time_step(message: str, verbosity: int = 0, min_verbosity: int = 1)[source]

Saves the time difference between last and current step.

Parameters:

  • message (str) – Message, that will be printed and saved, along with runtime.

  • verbosity (int, optional) – Current verbosity., defaults to 0

  • min_verbosity (int, optional) – If verbosity >= min_verbosity, print message., defaults to 1

skripts.VAE.VAE_smac.validate_incumbent(incumbent, fascade, run_dir: str, verbosity: int = 0)[source]

Saves the history of one run.

Parameters:

  • incumbent (Union[list, dict, ConfigSpace.Configuration]) – The calculated incumbent (best hyperparameters)

  • fascade (smac.AbstractFacade) – The fascade used for computation

  • run_dir (str) – Directory for saving the run.

  • verbosity (int, optional) – Level of verbosity, defaults to 0

Returns:

Best Hyperparameters

Return type:

ConfigSpace.Configuration

skripts.VAE.smac_runhistories module

skripts.VAE.vae module

Module contents