v0.3.1

• add memory to lookup score of previous evals to save time
• add hyperband initialization

v0.3.0

• add evolution strategy optimizer
• improvment of neighbour selection of simulated annealing
• bug fix in random search

v0.2.3

• restructuring of base-, search_space- and model-class
• cleanup of all optimizers
• using numpy arrays for positions instead of dict
• confining the movement in searchspace is done with numpy clip
• information flow now similar in all optimizers

v0.2.2

• restructuring of model evaluations
• bug fix for metrics of warm start points

v0.2.1

• enable transfer learning

v0.2.0

• improved start point functionality
• support for every metric listed in sklearn and keras doc
• multiple bug fixes and stability improvements

v0.1.4

• keras parameters for compile and fit are now passed by search_dict
• some stability improvements

v0.1.3

• XGBoost models can be optimized
• Keras models can be optimized (n_jobs are set to 1)
• fix number of steps
• change default values
• reconstruction of base class

v0.1.2

• add random_state
• code cleanup in simulated annealing
• add verbosity
• add start point

v0.1.1

• add score and export methods
• bug fix in PSO