In practice when comparing algorithms, the theoretical computational complexity can become less useful, especially when it comes to degenerate or already half solved cases. An algorithm that might have a better computational complexity can in some cases be better than an algorithm with a worse complexity.
To help a researcher investigate algorithms within the space of their input parameters, three procedures were developed and then implemented within the ALGator system. The first and second procedure tackle finding one or many parameter limits given a maximum execution time and the third procedure attempts to find a function which describes the time complexity given one free parameter. The procedures provide an automated starting analysis, from which a researcher can continue, while aiming to reduce the amount of algorithm executions needed while maximizing accuracy.
|
