SciPy parameter name | Description | Value | Comment |
---|---|---|---|
Basin-Hopping | |||
 T | Temperature | 0.8 | Changed to match the values used by Wales and Doye [67] |
 stepsize | Maximum step in each dimension that can be taken by the random displacement | 1 |  |
 niter | Number of Monte Carlo steps and local optimizations | 100 | 5000 was used by Wales and Doye [67] but their results show that global minima were often found in the first few hundred iterations. Since we are not interested in actually obtaining the global minimum, we select a value of 100 to make the cost of the optimizations bearable. This is sufficiently long in lower dimensions, to locate the global minimum, and sufficiently long in higher dimensions to make a fair comparison of performance |
Dual Annealing | |||
 initial_temp | Initial temperature | 50000 | Governs the maximum step the random displacement can take. Increased from the default to make the optimizer more exploratory since early test work showed a propensity to get stuck in the first minimum located |
 restart_temp_ratio | Ratio between current and initial temperatures which resets the temperature to the initial value | 0.01 | Increased from default to actually trigger new restarts and force the optimizer to explore other minima |