4. Summary and perspectives
Deep learning is currently the most active topic of machine learning. In recent years, deep learning models have been widely used in many fields such as computer vision, health monitoring and natural language processing. The explosion of big data offers enough training objects, which helps to improve the performance of deep learning. Furthermore, high-performance computing devices and architectures such as graphic processing units and CPU cluster enable the training of large-scale deep learning models for big data feature leaning. Today, deep learning models enjoy the success with a great many parameters, typically millions of parameters, together with a large number of training objects. While big data brings enough training objects, it also poses some challenges on deep learning. Therefore, in the past few years, many deep learning models have been developed for big data learning. In this paper, we provide a survey of big data deep learning models. Big data is typically defined by the four V’s model: volume, variety, velocity and veracity, which implies huge amount of data, various types of data, real-time data and low-quality data, respectively. Therefore, we summarized the deep learning models for big data learning from four aspects accordingly. In detail, we reviewed large-scale deep learning models for huge amount of data, multi-modal deep learning models and deep computation models for heterogeneous data, incremental deep learning models for real-time data, and reliable deep learning models for low-quality data. From the previous studies, we can see that deep learning models have made a great progress in big data feature learning. However, big data deep learning is still in its infancy, i.e., there are still some remaining challenges to be addressed for big data learning.
