Image Recognition with MapReduce Based Convolutional Neural Networks

Convolutional neural networks (CNNs) have gained global recognition in advancing the field of artificial intelligence and have had great successes in a wide array of applications including computer vision, speech and natural language processing. However, due to the rise of big data and increased complexity of tasks, the efficiency of training CNNs have been severely impacted. To achieve state-of-art results, CNNs require tens to hundreds of millions of parameters that need to be fine-tuned, resulting in extensive training time and high computational cost. To overcome these obstacles, this work takes advantage of distributed frameworks and cloud computing to develop a parallel CNN algorithm. Close examination of the implementation of MapReduce based CNNs as well as how the proposed algorithm accelerates learning are discussed and demonstrated through experiments. Results reveal high accuracy in classification and improvements in speedup, scaleup and sizeup compared to the standard algorithm.