In the era of "big data", it is becoming more of a challenge to not only build state-of-the-art predictive models, but also gain an understanding of what's really going on in the data. For example, it is often of interest to know which, if any, of the predictors in a fitted model are relatively influential on the predicted outcome. Some modern algorithms-like random forests and gradient boosted decision trees-have a natural way of quantifying the importance or relative influence of each feature. Other algorithms-like naive Bayes classifiers and support vector machines-are not capable of doing so and model-free approaches are generally used to measure each predictor's importance. In this paper, we propose a standardized, model-based approach to measuring predictor importance across the growing spectrum of supervised learning algorithms. Our proposed method is illustrated through both simulated and real data examples. The R code to reproduce all of the figures in this paper is available in the supplementary materials.

We propose a framework for the deployment We believe that the initial deployment of an application and its subsequent evolution in the face of host failures and other perturbations are separate but closely related problems. Both are too complex in large applications to be handled by a human operator. We propose that both should be controlled automatically, driven by a high-level configuration goal specified by the administrator at the outset. We thus address specifically the first and third of Kephart & Chess' issues [1]: self-configuration and self-healing.Our general approach is shown below. The application administrator specifies a deployment goal in terms of resources available and constraints over their deployment. We propose a new domain-specific constraint language called Deladas (DEclarative LAnguage for Describing Autonomic Systems) for this purpose. The resources include software components and physical hosts on which these components may be installed and executed. Constraints operate over aspects such as the mapping of components to hosts and the interconnection topology between components.The autonomic cycle is controlled by an engine, which we call the Autonomic Deployment and Management Engine (ADME), that attempts to satisfy a goal, specified by the administrator in the constraint language. The engine includes a parser and constraint solver. The result of the attempted goal satisfaction is a set of zero or more solutions. Each solution is in the form of a configuration, expressed as a Deployment Description Document (DDD), which describes a particular mapping of components to hosts and interconnection topology that satisfies the constraints.If a configuration can be found, it is enacted by the engine to produce a running deployment of the application. From a DDD, the ADME generates a collection of scripts which perform installation, instantiation and wiring of the components. Once the scripts have executed on the appropriate hosts, the application is fully deployed in its initial configuration [2].

A GLObal Smart Space (GLOSS) provides support for interaction amongst people, artefacts and places while taking account of both context and movement on a global scale. Crucial to the definition of a GLOSS is the provision of a set of location-aware services that detect, convey, store and exploit location information. We use one of these services, hearsay, to illustrate the implementation dimensions of a GLOSS. The focus of the paper is on both local and global software architecture to support the implementation of such services. The local architecture is based on XML pipe-lines and is used to construct location-aware components. The global architecture is based on a hybrid peer-to-peer routing scheme and provides the local architectures with the means to communicate in the global context.Comment: 4th International Conference on Mobile Data Management (MDM 2003

This paper describes an implemented system that is designed to support the deployment of applications offering distributed services, comprising a number of distributed components. This is achieved by creating high level placement and topology descriptions that drive tools to deploy applications consisting of components running on multiple hosts. The system addresses issues of heterogeneity by providing abstractions over host-specific attributes yielding a homogeneous run-time environment into which components may be deployed. The run-time environments provide secure binding mechanisms that permit deployed components to bind to stored data and services on the hosts on which they are running.

The hog deer Axis porcinus was believed to be extinct in Sri Lanka until a recent preliminary survey revealed a single remaining population in a 35-sq-km area of the south-western coastal belt. The authors describe the results of their 3-month study, outline potential threats to the survival of the population and discuss possible conservation measures.