AutoML

Created: 2023-02-13 14:46
#note

AutoML is the automation of machine learning activities to enable modeling building and decision-making without human intervention throughout the ML life cycle.

Scope:

• feature engineering -> delete useless feature, feature correlation and selection;
• model training -> model selection, hyperparameter tuning, ensemble training;
• deployment;
• customization -> model customization by use case.

Benefits:

• efficiency;
• speed of training and deployment;
• better results (search of optimal hyperparameters);
• accessibility.

Shortcomings:

• new use cases may create errors;
• bias may go untracked;
• processing capacity;
• explainability.

Tags

#mlops #ml