Many of today's data-intensive problems, such as data mining and machine learning, push the boundaries of conventional computing architectures with ever-increasing requirements for greater performance. GPUs are a good match for these applications because of their high memory bandwidth and massive computation power. But achieving high performance for this class of applications using GPUs can be quite challenging because of irregular data access patterns and complex heuristics employed in many of the data processing algorithms. Only a few recent efforts have resulted in productive GPU implementations of data-intensive applications, some of which are included in this section.

The use of GPUs in data-intensive applications is poised to explode in the near future. Although in the past many scientific communities focused on how to obtain more experimental or observed data, today scientists are concerned about what to do with the flood of data produced by modern scientific instruments. Fast analysis of very large volumes of data is now of paramount importance in astronomy, biology, finance, and medicine, and software developers are more and more inclined to take advantage of the GPU hardware to achieve the desirable level of performance.

Chapter 19 , written by Jerod J. Weinman, Augustus Lidaka, and Shitanshu Aggarwal, describes a discriminative maximum entropy learning algorithm, a machine-learning technique that builds a probability model from known data to make predictions about previously unseen data. CUBLAS library is used for one of the stages; other stages involve finding maximum value in an array for which the authors present two solutions. The program is implemented as a back end to Matlab.

Chapter 20 , written by Sergio Herrero-Lopez, describes the support vector machine, a supervised learning technique for classification and regression. The sequential minimal optimization algorithm is used to solve the multiclass classification problem. The implementation is decomposed into map and reduce stages realized as separate GPU kernels. Multi-GPU cascaded implementation is also presented.

Chapter 21 , written by Paul Richmond and Daniela Romano, describes a large-scale agent-based simulation framework for modeling the behavior of complex interacting systems. The GPU code for modeling individual agents is autogenerated from XML templates using a finite state machine-based abstract model of an agent. The formulation of agents’ behavior as distinct states reduces thread divergence and results in simple memory access patterns for better memory access coalescing. Real-time agent visualization is also presented.

Chapter 22 , written by Robin Weiss, describes a GPU implementation of an ant colony optimization algorithm, a rule-based classification technique used to solve a range of hard optimization problems. Major simulation steps are mapped into separate GPU kernels, with each ant modeled by a single GPU thread.