Abstract
We describe GPU implementations of the matrix recommender algorithms CCD++ and ALS. We compare the processing time and predictive ability of the GPU implementations with existing multi- core versions of the same algorithms. Results on the GPU are better than the results of the multi- core versions (maximum speedup of 14.8).

Abstract
In the past two decades, grid computing have fostered advances in several scientific domains by making resources available to a wide community and bridging scientific gaps. Grid infrastructures have been harnessing computational resources all around the world allowing all kinds of parallelisms to be explored. Other approaches to parallel and distributed computing still exist like the use of dedicated high-performance (HPC) infrastructures, and the use of clouds for computing and storage, but grid computing continues to be the predominant technology used for scientific computing in Europe, through the European Grid Infrastructure (EGI) and the European Middleware Initiative (EMI). Currently, there is a trend towards the use of cloud technologies for computing and storage. In Europe, this trend is being followed by taking advantage of all the experiences gained from building grid infrastructures and the technologies developed around them (resource management orchestration, unified job description languages, security, user interfaces, programming models, and scheduling policies, among others). As a result, the European Grid Infrastructure Federated Cloud is being built on top of the grid infrastructure already available. After almost two decades of the development of grid software and components and the emergence of competing technologies, now is the time to discuss current trends and to assess future prospects. When organizing this book, the authors considered contributions that would review the current grid computing scenario as well as contributions that would summarize the main tools and technologies used so far. The chapters in this book provide reviews for the following topics: a) performance prediction for parallel and distributed computing systems, b) resource sharing on computational grids, c) economic models for resource management, and d) programming frameworks. The chapters address grid issues such as a) the challenges of designing efficient job schedulers for production grids, b) scalability analysis of bag-of-tasks applications, c) the energy efficiency of resource reservation-based scheduling, and d) the development of parallel applications using the grid environment. Additionally, the following tools are presented: a) a programming framework based on the concept of a pluggable grid service that avoids explicit calls to grid services in scientific code and b) a desktop grid framework that runs on top of a cloud and can be deployed on the fly. The authors were each invited to contribute a chapter to this book, which were carefully revised and selected based on their originality and the value of their contribution to the overall discussion on grid computing and its future prospects.

Abstract

Abstract
Bipolar Disorder (BD) is a chronic and disabling disease that usually appears around 20 to 30 years old. Patients who suffer with BD may struggle for years to achieve a correct diagnosis, and only 50% of them generally receive adequate treatment. In this work we apply a machine learning technique called Inductive Logic Programming (ILP) in order to model relapse and no-relapse patients in a first attempt in this area to improve diagnosis and optimize psychiatrists' time spent with patients. We use ILP because it is well suited for our multi-relational dataset and because a human can easily interpret the logical rules produced. Our classifiers can predict relapse cases with 92% Recall and no-relapse cases with 73% Recall. The rules and variable theories generated by ILP reproduce some findings from the scientific literature. The generated multi-relational models can be directly interpreted by clinicians and researchers, and also open space to research biological mechanisms and interventions.

Abstract

Abstract
The development and use of computerized decision-support systems in the domain of breast cancer has the potential to facilitate the early detection of disease as well as spare healthy women unnecessary interventions. Despite encouraging trends, there is much room for improvement in the capabilities of such systems to further alleviate the burden of breast cancer. One of the main challenges that current systems face is integrating and translating multi-scale variables like patient risk factors and imaging features into complex management recommendations that would supplement and/or generalize similar activities provided by subspecialty-trained clinicians currently. In this chapter, we discuss the main types of knowledge-objectattribute, spatial, temporal and hierarchical-present in the domain of breast image analysis and their formal representation using two popular techniques from artificial intelligence-Bayesian networks and first-order logic. In particular, we demonstrate (i) the explicit representation of uncertain relationships between low-level image features and high-level image findings (e.g., mass, microcalcifications) by probability distributions in Bayesian networks, and (ii) the expressive power of logic to generally represent the dynamic number of objects in the domain. By concrete examples with patient data we show the practical application of both formalisms and their potential for use in decision-support systems.