A Model-Based Approach for Making Ecological Inference from Distance Sampling Data

We consider a fully model-based approach for the analysis of distance sampling data. Distance sampling has been widely used to estimate abundance (or density) of animals or plants in a spatially explicit study area. There is, however, no readily available method of making statistical inference on the relationships between abundance and environmental covariates. Spatial Poisson process likelihoods can be used to simultaneously estimate detection and intensity parameters by modeling distance sampling data as a thinned spatial point process. A model-based spatial approach to distance sampling data has three main benefits: it allows complex and opportunistic transect designs to be employed, it allows estimation of abundance in small subregions, and it provides a framework to assess the effects of habitat or experimental manipulation on density. We demonstrate the model-based methodology with a small simulation study and analysis of the Dubbo weed data set. In addition, a simple ad hoc method for handling overdispersion is also proposed. The simulation study showed that the model-based approach compared favorably to conventional distance sampling methods for abundance estimation. In addition, the overdispersion correction performed adequately when the number of transects was high. Analysis of the Dubbo data set indicated a transect effect on abundance via Akaike’s information criterion model selection. Further goodness-of-fit analysis, however, indicated some potential confounding of intensity with the detection function.

Bandit-Based Algorithms for Budgeted Learning

We explore the problem of budgeted machine learning, in which the learning algorithm has free access to the training examples’ labels but has to pay for each attribute that is specified. This learning model is appropriate in many areas, including medical applications. We present new algorithms for choosing which attributes to purchase of which examples in the budgeted learning model based on algorithms for the multi-armed bandit problem. All of our approaches outperformed the current state of the art. Furthermore, we present a new means for selecting an example to purchase after the attribute is selected, instead of selecting an example uniformly at random, which is typically done. Our new example selection method improved performance of all the algorithms we tested, both ours and those in the literature.

Designing Software Product Lines for Testability

Software product line (SPL) engineering offers several advantages in the development of families of software products such as reduced costs, high quality and a short time to market. A software product line is a set of software intensive systems, each of which shares a common core set of functionalities, but also differs from the other products through customization tailored to fit the needs of individual groups of customers. The differences between products within the family are well-understood and organized into a feature model that represents the variability of the SPL. Products can then be built by generating and composing features described in the feature model. Testing of software product lines has become a bottleneck in the SPL development lifecycle, since many of the techniques used in their testing have been borrowed from traditional software testing and do not directly take advantage of the similarities between products. This limits the overall gains that can be achieved in SPL engineering. Recent work proposed by both industry and the research community for improving SPL testing has begun to consider this problem, but there is still a need for better testing techniques that are tailored to SPL development. In this thesis, I make two primary contributions to software product line testing. First I propose a new definition for testability of SPLs that is based on the ability to re-use test cases between products without a loss of fault detection effectiveness. I build on this idea to identify elements of the feature model that contribute positively and/or negatively towards SPL testability. Second, I provide a graph based testing approach called the FIG Basis Path method that selects products and features for testing based on a feature dependency graph. This method should increase our ability to re-use results of test cases across successive products in the family and reduce testing effort. I report the results of a case study involving several non-trivial SPLs and show that for these objects, the FIG Basis Path method is as effective as testing all products, but requires us to test no more than 24% of the products in the SPL.