AutoML

AutoML = Automated Machine Learning

Benefits

• reduce time-to-market
• lower ML barrier to entry for non-data scientists
• iterate quickly using ML & automation
• save scarce resources for more challenging use cases

AutoML workflow

• AutoML can automate this workflow:
  • problem identification -> algorithm selection -> data preprocessing -> hyperparameter tuning
• considerations
  • transparency and control

Automated model tuning

• popular algorithms
  • grid search
  • random search
  • Bayesian optimization
  • hyperband: based on bandit approach, which typically uses a combination of exploitation and exploration to find the best possible hyperparameters.

Tools

• Amazon SageMaker Autopilot: AWS for Data Science#^a5ba5d