Search for: All records

The learning speed of feed-forward neural networks is notoriously slow and has presented a bottleneck in deep learning applications for several decades. For instance, gradient-based learning algorithms, which are used extensively to train neural networks, tend to work slowly when all of the network parameters must be iteratively tuned. To counter this, both researchers and practitioners have tried introducing randomness to reduce the learning requirement. Based on the original construction of Igelnik and Pao, single layer neural-networks with random input-to-hidden layer weights and biases have seen success in practice, but the necessary theoretical justification is lacking. In this study, we begin to fill this theoretical gap. We then extend this result to the non-asymptotic setting using a concentration inequality for Monte-Carlo integral approximations. We provide a (corrected) rigorous proof that the Igelnik and Pao construction is a universal approximator for continuous functions on compact domains, with approximation error squared decaying asymptotically likeO(1/n) for the numbernof network nodes. We then extend this result to the non-asymptotic setting, proving that one can achieve any desired approximation error with high probability providednis sufficiently large. We further adapt this randomized neural network architecture to approximate functions on smooth, compact submanifolds of Euclidean space, providing theoretical guarantees in both the asymptotic and non-asymptotic forms. Finally, we illustrate our results on manifolds with numerical experiments. 

Virtual memory, specifically paging, is undergoing significant innovation due to being challenged by new demands from modern workloads. Recent work has demonstrated an alternative software only design that can result in simplified hardware requirements, even supporting purely physical addressing. While we have made the case for this Compiler- And Runtime-based Address Translation (CARAT) concept, its evaluation was based on a user-level prototype. We now report on incorporating CARAT into a kernel, forming Compiler- And Runtime-based Address Translation for CollAborative Kernel Environments (CARAT CAKE). In our implementation, a Linux-compatible x64 process abstraction can be based either on CARAT CAKE, or on a sophisticated paging implementation. Implementing CARAT CAKE involves kernel changes and compiler optimizations/transformations that must work on all code in the system, including kernel code. We evaluate CARAT CAKE in comparison with paging and find that CARAT CAKE is able to achieve the functionality of paging (protection, mapping, and movement properties) with minimal overhead. In turn, CARAT CAKE allows significant new benefits for systems including energy savings, larger L1 caches, and arbitrary granularity memory management. 

OpenMP implementations make increasing demands on the kernel. We take the next step and consider bringing OpenMP into the kernel. Our vision is that the entire OpenMP application, run-time system, and a kernel framework is interwoven to become the kernel, allowing the OpenMP implementation to take full advantage of the hardware in a custom manner. We compare and contrast three approaches to achieving this goal. The first, runtime in kernel (RTK), ports the OpenMP runtime to the kernel, allowing any kernel code to use OpenMP pragmas. The second, process in kernel (PIK) adds a specialized process abstraction for running user-level OpenMP code within the kernel. The third, custom compilation for kernel (CCK), compiles OpenMP into a form that leverages the kernel framework without any intermediaries. We describe the design and implementation of these approaches, and evaluate them using NAS and other benchmarks. 

Abstract With ongoing climate change, populations are expected to exhibit shifts in demographic performance that will alter where a species can persist. This presents unique challenges for managing plant populations and may require ongoing interventions, including in situ management or introduction into new locations. However, few studies have examined how climate change may affect plant demographic performance for a suite of species, or how effective management actions could be in mitigating climate change effects. Over the course of two experiments spanning 6 yr and four sites across a latitudinal gradient in the Pacific Northwest, United States, we manipulated temperature, precipitation, and disturbance intensity, and quantified effects on the demography of eight native annual prairie species. Each year we planted seeds and monitored germination, survival, and reproduction. We found that disturbance strongly influenced demographic performance and that seven of the eight species had increasingly poor performance with warmer conditions. Across species and sites, we observed 11% recruitment (the proportion of seeds planted that survived to reproduction) following high disturbance, but just 3.9% and 2.3% under intermediate and low disturbance, respectively. Moreover, mean seed production following high disturbance was often more than tenfold greater than under intermediate and low disturbance. Importantly, most species exhibited precipitous declines in their population growth rates (λ) under warmer‐than‐ambient experimental conditions and may require more frequent disturbance intervention to sustain populations.Aristida oligantha, a C4 grass, was the only species to have λ increase with warmer conditions. These results suggest that rising temperatures may cause many native annual plant species to decline, highlighting the urgency for adaptive management practices that facilitate their restoration or introduction to newly suitable locations. Frequent and intense disturbances are critical to reduce competitors and promote native annuals’ persistence, but even such efforts may prove futile under future climate regimes. 

Abstract Predicting species' range shifts under future climate is a central goal of conservation ecology. Studying populations within and beyond multiple species' current ranges can help identify whether demographic responses to climate change exhibit directionality, indicative of range shifts, and whether responses are uniform across a suite of species.We quantified the demographic responses of six native perennial prairie species planted within and, for two species, beyond their northern range limits to a 3‐year experimental manipulation of temperature and precipitation at three sites spanning a latitudinal climate gradient in the Pacific Northwest, USA. We estimated population growth rates (λ) using integral projection models and tested for opposing responses to climate in different demographic vital rates (demographic compensation).Where species successfully established reproductive populations, warming negatively affectedλat sites within species' current ranges. Contrarily, warming and drought positively affectedλfor the two species planted beyond their northern range limits. Most species failed to establish a reproductive population at one or more sites within their current ranges, due to extremely low germination and seedling survival. We found little evidence of demographic compensation buffering populations to the climate treatments.Synthesis. These results support predictions across a suite of species that ranges will need to shift with climate change as populations within current ranges become increasingly vulnerable to decline. Species capable of dispersing beyond their leading edges may be more likely to persist, as our evidence suggests that projected changes in climate may benefit such populations. If species are unable to disperse to new habitat on their own, assisted migration may need to be considered to prevent the widespread loss of vulnerable species.