The Nature of Statistical Learning Theory (Vapnik)

Details

Photo

—

Name

—

Field

Machine learning / Statistics

Year

1995

Core idea in one line

There’s a disciplined way to think about learning from finite data without lying to yourself about generalization.

My take

Reads like the stern but fair teacher behind half of modern ML; not subway-light reading, but a good way to humble your inner data wizard.

Good starting point for

Understanding why models overfit and what “capacity” actually means beyond buzzwords.