Nelder–Mead Simplex Search and Its Modified Optimization Variant

Summary

The article introduces Nelder–Mead as a derivative-free method for minimizing an objective function by moving a simplex through parameter space. It describes sorting simplex vertices by objective value, calculating the centroid excluding the worst vertex, and applying reflection, expansion, or contraction to replace that vertex. The author also discusses shrinkage, which contracts all vertices toward the best point, but omits it in the modified implementation because of its computational cost and concern that it may collapse exploration.

The proposed modification uses multiple simplex agents and Levy flights in an attempt to escape local optima. The article reports that the method behaves acceptably on smooth functions with few variables but struggles with more complex, discrete optimization problems such as trading parameter searches. Its assessment lists low speed, poor scalability, result variability, and a tendency to become trapped in local extrema. These conclusions are based on the author’s experiments; the article notes that implementations were modified and cautions that its descriptions may not exactly match canonical algorithms.

Key ideas

• Nelder–Mead searches without derivatives by replacing the worst simplex vertex through reflection, expansion, or contraction.
• The author omits shrinkage in the modified version, citing its computational expense and risk of ending exploration near one local extremum.
• Multiple simplex agents and Levy flights are introduced as experimental measures to help escape local optima.
• The reported experiments favor smooth objectives with few variables and show difficulties on complex, discrete trading optimization problems.
• The modified algorithm is reported to be slow, poorly scalable, variable in outcome, and prone to local traps.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.