University of California, Santa Cruz

About: University of California, Santa Cruz is a education organization based out in Santa Cruz, California, United States. It is known for research contribution in the topics: Galaxy & Population. The organization has 15541 authors who have published 44120 publications receiving 2759983 citations. The organization is also known as: UCSC & UC, Santa Cruz.

Exploiting Generative Models in Discriminative Classifiers

Abstract: Generative probability models such as hidden Markov models provide a principled way of treating missing information and dealing with variable length sequences. On the other hand, discriminative methods such as support vector machines enable us to construct flexible decision boundaries and often result in classification performance superior to that of the model based approaches. An ideal classifier should combine these two complementary approaches. In this paper, we develop a natural way of achieving this combination by deriving kernel functions for use in discriminative methods such as support vector machines from generative probability models. We provide a theoretical justification for this combination as well as demonstrate a substantial improvement in the classification performance in the context of DNA and protein sequence analysis.

Founding events in species invasions: genetic variation, adaptive evolution, and the role of multiple introductions

Abstract: Invasive species are predicted to suffer from reductions in genetic diversity during founding events, reducing adaptive potential. Integrating evidence from two literature reviews and two case studies, we address the following questions: How much genetic diversity is lost in invasions? Do multiple introductions ameliorate this loss? Is there evidence for loss of diversity in quantitative traits? Do invaders that have experienced strong bottlenecks show adaptive evolution? How do multiple introductions influence adaptation on a landscape scale? We reviewed studies of 80 species of animals, plants, and fungi that quantified nuclear molecular diversity within introduced and source populations. Overall, there were significant losses of both allelic richness and heterozygosity in introduced populations, and large gains in diversity were rare. Evidence for multiple introductions was associated with increased diversity, and allelic variation appeared to increase over long timescales (~100 years), suggesting a role for gene flow in augmenting diversity over the long-term. We then reviewed the literature on quantitative trait diversity and found that broad-sense variation rarely declines in introductions, but direct comparisons of additive variance were lacking. Our studies of Hypericum canariense invasions illustrate how populations with diminished diversity may still evolve rapidly. Given the prevalence of genetic bottlenecks in successful invading populations and the potential for adaptive evolution in quantitative traits, we suggest that the disadvantages associated with founding events may have been overstated. However, our work on the successful invader Verbascum thapsus illustrates how multiple introductions may take time to commingle, instead persisting as a 'mosaic of maladaptation' where traits are not distributed in a pattern consistent with adaptation. We conclude that management limiting gene flow among introduced populations may reduce adaptive potential but is unlikely to prevent expansion or the evolution of novel invasive behaviour.

Kelp forest ecosystems: biodiversity, stability, resilience and future

Abstract: Kelp forests are phyletically diverse, structurally complex and highly productive components of coldwater rocky marine coastlines. This paper reviews the conditions in which kelp forests develop globally and where, why and at what rate they become deforested. The ecology and long archaeological history of kelp forests are examined through case studies from southern California, the Aleutian Islands and the western North Atlantic, well-studied locations that represent the widest possible range in kelp forest biodiversity. Global distribution of kelp forests is physiologically constrained by light at high latitudes and by nutrients, warm temperatures and other macrophytes at low latitudes. Within mid-latitude belts (roughly 40–60° latitude in both hemispheres) well-developed kelp forests are most threatened by herbivory, usually from sea urchins. Overfishing and extirpation of highly valued vertebrate apex predators often triggered herbivore population increases, leading to widespread kelp deforestation. Such deforestations have the most profound and lasting impacts on species-depauperate systems, such as those in Alaska and the western North Atlantic. Globally urchin-induced deforestation has been increasing over the past 2–3 decades. Continued fishing down of coastal food webs has resulted in shifting harvesting targets from apex predators to their invertebrate prey, including kelp-grazing herbivores. The recent global expansion of sea urchin harvesting has led to the widespread extirpation of this herbivore, and kelp forests have returned in some locations but, for the first time, these forests are devoid of vertebrate apex predators. In the western North Atlantic, large predatory crabs have recently filled this void and they have become the new apex predator in this system. Similar shifts from fish- to crab-dominance may have occurred in coastal zones of the United Kingdom and Japan, where large predatory finfish were extirpated long ago. Three North American case studies of kelp forests were examined to determine their long history with humans and project the status of future kelp forests to the year 2025. Fishing impacts on kelp forest systems have been both profound and much longer in duration than previously thought. Archaeological data suggest that coastal peoples exploited kelp forest organisms for thousands of years, occasionally resulting in localized losses of apex predators, outbreaks of sea urchin populations and probably small-scale deforestation. Over the past two centuries, commercial exploitation for export led to the extirpation of sea urchin predators, such as the sea otter in the North Pacific and predatory fishes like the cod in the North Atlantic. The large-scale removal of predators for export markets increased sea urchin abundances and promoted the decline of kelp forests over vast areas. Despite southern California having one of the longest known associations with coastal kelp forests, widespread deforestation is rare. It is possible that functional redundancies among predators and herbivores make this most diverse system most stable. Such biodiverse kelp forests may also resist invasion from non-native species. In the species-depauperate western North Atlantic, introduced algal competitors carpet the benthos and threaten future kelp dominance. There, other non-native herbivores and predators have become established and dominant components of this system. Climate changes have had measurable impacts on kelp forest ecosystems and efforts to control the emission of greenhouse gasses should be a global priority. However, overfishing appears to be the greatest manageable threat to kelp forest ecosystems over the 2025 time horizon. Management should focus on minimizing fishing impacts and restoring populations of functionally important species in these systems.

The Eighth Data Release of the Sloan Digital Sky Survey: First Data from SDSS-III

Abstract: The Sloan Digital Sky Survey (SDSS) started a new phase in 2008 August, with new instrumentation and new surveys focused on Galactic structure and chemical evolution, measurements of the baryon oscillation feature in the clustering of galaxies and the quasar Lyα forest, and a radial velocity search for planets around ~8000 stars. This paper describes the first data release of SDSS-III (and the eighth counting from the beginning of the SDSS). The release includes five-band imaging of roughly 5200 deg2 in the southern Galactic cap, bringing the total footprint of the SDSS imaging to 14,555 deg2, or over a third of the Celestial Sphere. All the imaging data have been reprocessed with an improved sky-subtraction algorithm and a final, self-consistent photometric recalibration and flat-field determination. This release also includes all data from the second phase of the Sloan Extension for Galactic Understanding and Exploration (SEGUE-2), consisting of spectroscopy of approximately 118,000 stars at both high and low Galactic latitudes. All the more than half a million stellar spectra obtained with the SDSS spectrograph have been reprocessed through an improved stellar parameter pipeline, which has better determination of metallicity for high-metallicity stars.

A Simple Weight Decay Can Improve Generalization

Abstract: It has been observed in numerical simulations that a weight decay can improve generalization in a feed-forward neural network. This paper explains why. It is proven that a weight decay has two effects in a linear network. First, it suppresses any irrelevant components of the weight vector by choosing the smallest vector that solves the learning problem. Second, if the size is chosen right, a weight decay can suppress some of the effects of static noise on the targets, which improves generalization quite a lot. It is then shown how to extend these results to networks with hidden layers and non-linear units. Finally the theory is confirmed by some numerical simulations using the data from NetTalk.