In shortLearning from examples / Training and testing / Bias and fairness / Small models, big value.
1. Learning from examples
A model sees many examples and adjusts until it can predict new ones. It is like a hen learning where the feed usually appears.
2. Training and testing
Keep some data aside to test the model. If it only works on training data, it has memorized, not learned.
3. Bias and fairness
A model reflects its data. If the data is skewed, so is the model. Check, measure, and correct.
4. Small models, big value
A modest model that runs everywhere often beats a giant one that runs nowhere.
Written by Beatrice Featherstone, Head of Site Reliability. Disagree? Tell us. We read every message, usually before the second coffee.