Procedure for Creating a Virtual Multibank Agent: Process
The tabu restrictions are not inviolable under any circumstance. When a tabu movement provides a better solution than any other previously found, its classification as tabu can be eliminated. The condition that enables such elimination is called aspiration criteria. Finally, we will establish finalization criteria based on which a maximum number of iterations can be established, or a maximum number of steps without improving the previous solution obtained. The iterations are presented in tables, where the number of iteration of the process is indicated, the itineraries generated and the corresponding values of the objective function, as well as the itinerary proposed for the tabu movement and the better movement acceptable in order to continue the process.
The complete neighborhoods are examined, and the best change is made when obtaining the minimum value in the objective function and it is not a tabu movement or, in the case where it is a tabu movement, it must satisfy the aspiration criteria in order to be acceptable. For this case, the aspiration criteria are fulfilled if the value of the objective function improves on all the values previously found.
The matrix of frequencies enables recording the “history” of the procedure and that is what is used for forming the long-term memory function, thus enabling the diversification of the search as well as the possibility of directing the search towards “nearest” or “furthest” of the regions explored.
Let us suppose that in a navigation session at least two admissible movements are recorded which provide us the best values of the objective function, and therefore any one of the two may be chosen; in this case, the first is chosen, and the itinerary chosen would act as a starting point for the following iteration. In the case where itineraries with tabu movements are obtained in an iteration, but these satisfy the aspiration criteria, the itinerary with the smallest value in the neighborhood would be taken as the starting point for the next iteration. review
In general, if there are not any admissible points in a given iteration and the stopping criteria has not been satisfied, the mid-term (intensification) and long-term (diversification) memory functions would have to be used in order to continue with the search.
Normally, the search is restarted based on the best current solution. Some options would be that of maintaining set components or attributes that appear to be better or to change the neighborhood scheme.
When navigating on the Internet, one of the main goals is to make sure that the search for information is as efficient as possible, and that implies reducing the search time, the elimination of unproductive routines in the search, or information that is not useful or sufficient for the navigator. The adaptive memory of the tabu algorithm exploits the history of the process of resolving the problem by making reference to four main dimensions consisting of the property of being recent, frequency, quality and influence. All of these properties make it a method that is especially useful in the most efficient search for information.