Improving Geospatial Big Data Analytics Approaches

The advent of geospatial big data has led to a paradigm shift where most related applications became data driven, and therefore intensive in both data and computation. This revolution has covered most domains, namely the real-time systems such as web search engines, social networks, and tracking systems. These later are linked to the high-velocity feature, which characterizes the dynamism, the fast changing and moving data streams. Therefore, the response time and speed of such queries, along with the space complexity, are among data stream analysis system requirements, which still require improvements using sophisticated algorithms. In this vein, this chapter discusses new approaches that can reduce the complexity and costs in time and space while improving the efficiency and quality of responses of geospatial big data stream analysis to efficiently detect changes over time, conclude, and predict future events.

On Training Deep Neural Networks Using a Streaming Approach

In recent years, many deep learning methods, allowed for a significant improvement of systems based on artificial intelligence methods. Their effectiveness results from an ability to analyze large labeled datasets. The price for such high accuracy is the long training time, necessary to process such large amounts of data. On the other hand, along with the increase in the number of collected data, the field of data stream analysis was developed. It enables to process data immediately, with no need to store them. In this work, we decided to take advantage of the benefits of data streaming in order to accelerate the training of deep neural networks. The work includes an analysis of two approaches to network learning, presented on the background of traditional stochastic and batch-based methods.