An Extended Kernel for Generalized Multiple-Instance Learning

The multiple-instance learning (MIL) model has been successful in areas such as drug discovery and content-based image-retrieval. Recently, this model was generalized and a corresponding kernel was introduced to learn generalized MIL concepts with a support vector machine. While this kernel enjoyed empirical success, it has limitations in its representation. We extend this kernel by enriching its representation and empirically evaluate our new kernel on data from content-based image retrieval, biological sequence analysis, and drug discovery. We found that our new kernel generalized noticeably better than the old one in content-based image retrieval and biological sequence analysis and was slightly better or even with the old kernel in the other applications, showing that an SVM using this kernel does not overfit despite its richer representation.

Distribution of Dioecious Eastern Red Cedar (Juniperus virginiana) along an Environmental Gradient in Ogallala, NE.

Abstract The purpose of this research was to study the sex distribution and energy allocation of dioecious Eastern Red Cedars (Juniperus virginiana) along an environmental resource gradient. The trees surveyed were growing in a canyon located at the University of Nebraska’s Cedar Point Biological Research Station in Ogallala, Nebraska. Due to the geography of this canyon, environmental factors necessary for plant growth should vary depending on the tree’s location within the canyon. These factors include water availability, sun exposure, ground slope, and soil nitrogen content, all of which are necessary for carbon acquisition. Juniperus virginiana is a dioecious conifer. Dioecious plants maintain male and female reproductive structures on separate individuals. Therefore, proximal spatial location is essential for pollination and successful reproduction. Typically female reproductive structures are more costly and require a greater investment of carbon and nitrogen. For this reason, growth, survival and successful reproduction are more likely to be limited by environmental resources for females than for male individuals. If this is true for Juniperus virginiana, females should be located in more nutrient and water rich areas than males. This also assumes that females can not be reproductively successful in areas of poor environmental quality. Therefore, reproductive males should be more likely to inhabit environments with relatively lower resource availability than females. Whether the environment affects sexual determination or just limits survival of different sexes is still relatively unknown. In order to view distribution trends along the environmental gradient, the position of the tree in the canyon transect was compared to its sex. Any trend in sex should correspond with varying environmental factors in the canyon, ie: sunlight availability, aspect, and ground slope. The individuals’ allocation to growth and reproduction was quantified first by comparing trunk diameter at six inches above ground to sex and location of the tree. The feature of energy allocation was further substantiated by comparing carbon and nitrogen content in tree leaf tissue and soil to location and sex of each individual. Carbon and nitrogen in soil indicate essential nutrient availability to the individual, while C and N in leaf tissue indicate nutrient limitation experienced by the tree. At the conclusion of this experiment, there is modest support that survival and fecundity of females demands environments relatively richer in nutrients, than needed by males to survive and be reproductively active. Side of the canyon appeared to have an influence on diameter of trees, frequency of sex and carbon and nitrogen leaf content. While this information indicated possible trends in the relation of sex to nutrient availability, most of the environmental variables presumed responsible for the sex distribution bias differed minutely and may not have been biologically significant to tree growth.