Good summary provided by pyhopper paper
- HyperOpt: parallelization is supported via Apache Spark
- Optuna: parallelization is supported via MySQL
- NeverGrad: developed by Facebook Research
- keras-tuner: have to add layers in a specific way, no parallelization support
- Autotune
- Dragonfly
- RayTune: Distributed processing, allows use of HyperOpt etc.
- PyHopper: Distributed, Intuitive, Lack of integrations as new package