High performance analysis in generating FTTM graph of pseudo degree zero using CPU-GPU

Fuzzy topological topographic mapping (FTTM) is developed to solve neuro magnetic inverse problem. Due to its special characteristics, namely the homeomorphism between its components, new elements can be generated using components of different version. The generated element can be depicted as a graph. A graph of pseudo degree zero is generated by FTTM with isolated component. The amount of data generated is categorized as a big data since the number of FTTM raise exponentially as the number of components rises. In this paper a sequential algorithm in generating FTTM of pseudo degree zero is presented. The results show the runtime of the algorithm also rises exponentially as the number of components increase. A GPU-CPU parallel algorithm is proposed to cut the runtime. This algorithm succeeded in increasing the speed up to 363% for data with large number of components but decrease the speed for data with low number of components. This slowdown is due to the communication time between the CPU and GPU.