Tutorials

• Search space constructs
  • Starting with a fixed Keras model
  • Introducing independent hyperparameters
  • Sharing hyperparameters across modules
  • Expressing dependencies between hyperparameters
  • Delaying sub-search space creation through substitution
  • Minor details
  • Concluding remarks
• Implementing new modules
  • Starting with the framework helper module
  • Working with inputs and outputs instead of modules
  • Simplifications for single-input single-output modules
  • Easily creating modules from framework functions
  • Implementing new substitution modules
  • An example of a complex substitution module
  • Concluding remarks
• Implementing new searchers
  • Searcher API
  • Random searcher
  • SMBO searcher
  • Concluding remarks
• Supporting new frameworks
  • Specializing the module class
  • Keras helpers
  • Pytorch helpers
  • Concluding remarks
• Logging and visualization
  • Starting Keras example to adapt for logging
  • Evaluator and search space definitions
  • Main search loop with logging
  • Concluding remarks