Hill climbing is a local search algorithm used in artificial intelligence to find optimal solutions by iteratively making incremental improvements. This article explores three major types: simple hill climbing, steepest ascent hill climbing, and stochastic hill climbing.
Simple Hill Climbing evaluates neighboring states one by one and moves to the first state that improves the current state. It is fast but can get stuck on local optima.
Steepest Ascent Hill Climbing examines all neighboring states and selects the one with the highest improvement. While it produces better results, it is computationally more expensive.
Stochastic Hill Climbing randomly selects among improving neighbors, with the probability of selection proportional to the improvement. This helps escape local optima at the cost of slower convergence.
All three algorithms share the same core limitation: they cannot guarantee finding a global optimum, but each offers trade-offs between speed, accuracy, and robustness.