Hill Climbing | Gradient Descent |
1. In Hill Climbing, you look at all neighboring states and evaluate the cost function in each of them. | 1. In Gradient Descent, you look at the slope of your local neighbor and move in the direction with the steepest slope. |
2. Hill Climbing is less efficient than Gradient Descent. | 2. Gradient Descent is much more efficient than Hill Climbing |
3. In Hill Climbing, you can optimize discrete problems, you just need to be able to evaluate a cost function for a given state. | 3. Gradient Descent assures you optimize a continuous function and you can compute it's gradient in a given state. |