Differences that Make a Difference: Comparing Selected Optimization Algorithms in R Language

Numerical optimization is very important for engineers, statisticians, as well as researchers and practitioners in other domains. The R language, which is at the top of recent popularity rankings, provides various packages and functions to perform optimization.
This paper compares several implementations of optimization methods that are available in R language. The results show that even implementations of the same method differ considerably from one another. The conducted experiments also reveal that several implementations have errors or some oddities in their source code which hinders or excludes their practical usage. Some of these errors can be bypassed or avoided which is also discussed within the paper. Finally it is suggested which implementation should be used to achieve reasonable results without excessive effort.