The Method of Ensembling Neural Networks for Pattern Recognition

Abstract

Today it is difficult to imagine creating various applications without an artificial intelligence. Its capabilities increase the fault tolerance of the system, give it variability in actions, allow it to respond to non-standard situations in accordance with standard rules, and most importantly. It allows the system to learn during operation on specific work processes. Image recognition algorithms can be used in many areas of human life: medicine, security, production process control, road traffic, the creation of unmanned vehicles, Internet searches. Neural network approaches to solving pattern recognition problems have shown their high efficiency and accuracy. Convolutional neural networks are currently the most effective in recognition. The principle of their operation is based on alternating convolutional and subsampling layers. There are also areas of human life where recognition accuracy is of fundamental importance, for example medicine, and in such areas it is advisable to focus on improving accuracy using various methods, for example, the ensemble method presented.