Dense vs Sparse Feature

  • Dense Feature: zeros are rare or non-existent.
  • Sparse Feature: non-zeros are rare

Q: When is an email not an email? A: When it’s a “feature vector”

Objects are represented as a vector of features:

  • Dense Features: sender IP, timestamp, # of recipients, etc.
  • Sparse Features: message contains “Viagra”, subject contains “URGENT”, etc.