Implementación de una modificación del algoritmo de Huffman utilizando redes neuronales

Abstract

In this work a compression algorithm is implemented, to be used in the process of coding the source of information in a telecommunications system. This implementation includes a modification in the data structure managed by the Huffman algorithm, which canonically uses a binary tree abstract data structure to distribute the characters or symbols of the information burst to be encoded, classifying them, depending on the frequencies relative of appearance. This tree can be replaced by a neural network that is similar in its topology to a directed k-complete graph, with weights in the connections or edges, in this way it is possible to train the neural network so that it finds patterns in the limited bursts of information in size, fragmenting the message and thus increasing the compression rate of the information to be sent to the channel, this result will also be framed in the analysis of the temporal complexity required by the algorithm to be executed; If it is possible to reduce the volume of information, it would mean an improvement in the degree of service and performance of the telecommunications network by increasing capacity without depending on the type of channel or its bandwidth.