CS5720 - Week 3
Slide 56 of 60

Grid Search vs Random Search

📏

Grid Search

Systematically evaluates all combinations of hyperparameter values from a predefined grid.

✅ Advantages
  • • Exhaustive coverage of search space
  • • Reproducible results
  • • Easy to implement and understand
  • • Guaranteed to find best within grid
❌ Disadvantages
  • • Computationally expensive
  • • Curse of dimensionality
  • • Wastes time on unimportant params
  • • Fixed discretization
🎲

Random Search

Randomly samples hyperparameter combinations from specified distributions.

✅ Advantages
  • • More efficient exploration
  • • Better for important parameters
  • • Can be stopped anytime
  • • Handles continuous parameters well
❌ Disadvantages
  • • No guarantee of coverage
  • • Results vary between runs
  • • May miss optimal regions
  • • Requires more iterations

Interactive Search Space Visualization

Grid Search
Random Search
Grid Points Evaluated
0
Random Points Evaluated
0
Grid Best Score
0.00
Random Best Score
0.00
← Previous Next →
Prepared by Dr. Gorkem Kar