Comparing deep learning performance on BigData by using CPUs and GPUs

During the last few years, we have witnessed a fast growth of the cyber world in which great attention is focused on the use of Big Data that cannot be managed or analyzed with the use of traditional tools. In the past, to convert raw data into valuable information, machine learning techniques were extensively preferred. However, conventional techniques such as Naive Bayes and support vector machines are not efficient for these huge data. Therefore, deep learning methods which are especially preferred in image/speech recognition, information retrieval, language translation, etc., arise as sophisticated and acceptable choices to process data with high accuracy in a hierarchical representation model. Depending on the size of data, even deep learning techniques can be inadequate. To increase the performance, some parallel processing techniques are needed which can be executed on a multi-core structure of the processor. With the technological improvements, the Graphical Processing Units(GPUs), which can contain a few thousands of cores, can also be used in this parallel execution platform. In this paper, we aimed to use a deep learning approach for processing big data to solve a specific problem in a multi-core platform. The experimental results are compared with a CPU execution, and it is depicted that use of GPU Technologies increases the performance of system up to 10 times depending on the type of the GPUs.