An analysis on sub-regional voting scheme.

Voting systems are a very intriguing subject. Different voting schemes have different stabilities in retaining the original results in the presence of noise(the environmental influence that changes voter's decision). A candidate could win an election under some voting schemes, even though he/she has less supporters than another candidate overall. This research introduces how sub-regional and regional voting systems change with the presence of concentrated noise. This paper presents a new model that analyzes the likelihood of sub-regional, regional and national changes in voting decisions under the influences of multiple concentrated-noise blocks. This model breaks down a nation into equal-sized regions, and a region into equal-sized sub-regions based on the idea of deep learning algorithms, which are distributed representations. In the model, sub-region, region and nation correspond to three levels of a hierarchy. An analysis on the voting results of sub-regions to obtain the voting result of the region is completed in this thesis. In the same way, the voting result of a nation can be obtained by analyzing the voting results of regions. This model utilizes the concept of deep learning' by using knowledge acquired from lower level representations to help define higher level concepts. --Leaf ii.The original print copy of this thesis may be available here: http://wizard.unbc.ca/record=b186319