The Time Complexity Analysis of Neural Network Model Configurations

Abstract

The neural network algorithms, such as the deep-learning approach, have been widely applied in dealing with the computer vision problems. The more sophisticated the neural network model is designed; the more computing resources and processing time will be consumed. The time-complexity analysis discloses the training and validating time processed versus the accuracy outcome of the neural network model. This paper proposes a rigorous framework of conducting the time-complexity analysis against the neural network models. From the experiment results against the progressively sophisticated neural network model design, the paper argues that pursuing an unreasonably high accurate result or in persisting in finding the perfect algorithm may not be worth and in practical from the time-complexity perspective if the data quality was not consistently in favor of the algorithm chosen.