Search for: All records

Distributed deep learning framework tools should aim at high efficiency of training and inference of distributed exascale deep learning algorithms. There are three major challenges in this endeavor: scalability, adaptivity and efficiency. Any future framework will need to be adaptively utilized for a variety of heterogeneous hardware and network environments and will thus be required to be capable of scaling from single compute node up to large clusters. Further, it should be efficiently integrated into popular frameworks such as TensorFlow, PyTorch, etc. This paper proposes a dynamically hybrid (hierarchy) distribution structure for distributed deep learning, taking advantage of flexible synchronization on both centralized and decentralized architectures, implementing multi-level fine-grain parallelism on distributed platforms. It is scalable as the number of compute nodes increases, and can also adapt to various compute abilities, memory structures and communication costs. 

Abstract As the Gulf Stream separates from the coast, it sheds both Warm and Cold Core Rings between $$75^\circ$$ 75 ∘ and $$55^\circ \,\hbox {W}$$ 55 ∘ W . We present evidence that this ring formation behavior has been asymmetric over both interannual and seasonal time-scales. After a previously reported regime-shift in 2000, 15 more Warm Core Rings have been forming yearly compared to 1980–1999. In contrast, there have been no changes in the annual formation rate of the Cold Core Rings. This increase in Warm Core Ring production leads to an excess heat transfer of 0.10 PW to the Slope Sea, amounting to 7.7–12.4% of the total Gulf Stream heat transport, or 5.4–7.3% of the global oceanic heat budget at $$30^\circ \,\hbox {N}$$ 30 ∘ N . Seasonally, more Cold Core Rings are produced in the winter and spring and more Warm Core Rings are produced in the summer and fall leading to more summertime heat transfer to the north of the Stream. The seasonal cycle of relative ring formation numbers is strongly correlated (r = 0.82) with that of the difference in upper layer temperatures between the Sargasso and Slope seas. This quantification motivates future efforts to understand the recent increasing influence of the Gulf Stream on the circulation and ecosystem in the western North Atlantic. 

Abstract Here, we implement a physics teaching laboratory practice that aims to investigate the Stokes parameters through the analysis of polarized light transmitted by polarizing optical elements. The practice was implemented according to the project-based learning methodology. All the investigated states of polarized light were generated using a homemade polarization simulator, comprising a light source (laser emitting at 670 nm) combined with different polarizing optical elements. We have shown that the practice is effective for obtaining the Stokes' parameters. Moreover, it can be easily designed as an advanced experimental project for optics teaching laboratories, with the purpose of confirming and advancing theoretical studies of the topic. 

Abstract Comprehensive sampling of the carbonate system in estuaries and coastal waters can be difficult and expensive because of the complex and heterogeneous nature of near-shore environments. We show that sample collection by community science programs is a viable strategy for expanding estuarine carbonate system monitoring and prioritizing regions for more targeted assessment. ‘Shell Day’ was a single-day regional water monitoring event coordinating coastal carbonate chemistry observations by 59 community science programs and seven research institutions in the northeastern United States, in which 410 total alkalinity (TA) samples from 86 stations were collected. Field replicates collected at both low and high tides had a mean standard deviation between replicates of 3.6 ± 0.3 µ mol kg −1 ( σ mean ± SE, n = 145) or 0.20 ± 0.02%. This level of precision demonstrates that with adequate protocols for sample collection, handling, storage, and analysis, community science programs are able to collect TA samples leading to high-quality analyses and data. Despite correlations between salinity, temperature, and TA observed at multiple spatial scales, empirical predictions of TA had relatively high root mean square error >48 µ mol kg −1 . Additionally, ten stations displayed tidal variability in TA that was not likely driven by low TA freshwater inputs. As such, TA cannot be predicted accurately from salinity using a single relationship across the northeastern US region, though predictions may be viable at more localized scales where consistent freshwater and seawater endmembers can be defined. There was a high degree of geographic heterogeneity in both mean and tidal variability in TA, and this single-day snapshot sampling identified three patterns driving variation in TA, with certain locations exhibiting increased risk of acidification. The success of Shell Day implies that similar community science based events could be conducted in other regions to not only expand understanding of the coastal carbonate system, but also provide a way to inventory monitoring assets, build partnerships with stakeholders, and expand education and outreach to a broader constituency. 

Abstract Survival of Gulf Stream (GS) warm core rings (WCRs) was investigated using a census consisting of a total of 961 rings formed during the period 1980–2017. Kaplan‐Meier survival probability and Cox hazard proportional models were used for the analysis. The survival analysis was performed for rings formed in four 5° zones between 75° W and 55° W. The radius, latitude, and distance from the shelf‐break of a WCR at formation all had a significant effect on the survival of WCRs. A pattern of higher survival was observed in WCRs formed in Zone 2 (70°–65° W) or Zone 3 (65°–60° W) and then demised in Zone 1 (75°–70° W). Survival probability of the WCRs increased to more than70%for those formed within a latitude band from 39.5° to 41.5° N. Survival probability is reduced when the WCRs are formed near the New England Seamounts. 

Human land use threatens global biodiversity and compromises multiple ecosystem functions critical to food production. Whether crop yield–related ecosystem services can be maintained by a few dominant species or rely on high richness remains unclear. Using a global database from 89 studies (with 1475 locations), we partition the relative importance of species richness, abundance, and dominance for pollination; biological pest control; and final yields in the context of ongoing land-use change. Pollinator and enemy richness directly supported ecosystem services in addition to and independent of abundance and dominance. Up to 50% of the negative effects of landscape simplification on ecosystem services was due to richness losses of service-providing organisms, with negative consequences for crop yields. Maintaining the biodiversity of ecosystem service providers is therefore vital to sustain the flow of key agroecosystem benefits to society.