Amazon.com

About: Amazon.com is a company organization based out in Seattle, Washington, United States. It is known for research contribution in the topics: Service (business) & Service provider. The organization has 13363 authors who have published 17317 publications receiving 266589 citations.

Energy-Based Learning for Scene Graph Generation

Abstract: Traditional scene graph generation methods are trained using cross-entropy losses that treat objects and relationships as independent entities. Such a formulation, however, ignores the structure in the output space, in an inherently structured prediction problem. In this work, we introduce a novel energy-based learning framework for generating scene graphs. The proposed formulation allows for efficiently incorporating the structure of scene graphs in the output space. This additional constraint in the learning framework acts as an inductive bias and allows models to learn efficiently from a small number of labels. We use the proposed energy-based framework† to train existing stateof-the-art models and obtain a significant performance improvement, of up to 21% and 27%, on the Visual Genome [9] and GQA [5] benchmark datasets, respectively. Furthermore, we showcase the learning efficiency of the proposed framework by demonstrating superior performance in the zero- and few-shot settings where data is scarce.

Dynamic scaling of a cluster of computing nodes used for distributed execution of a program

Abstract: Techniques are described for managing distributed execution of programs, including by dynamically scaling a cluster of multiple computing nodes used to perform ongoing distributed execution of a program, such as to increase and/or decrease the quantity of computing nodes in the cluster at various times and for various reasons. An architecture may be used that facilitates the dynamic scaling of a cluster, including by having at least some of the computing nodes act as core nodes that each participate in a distributed storage system for the distributed program execution, and having one or more other computing nodes that act as auxiliary nodes that do not participate in the distributed storage system. If computing nodes are selected to be removed from the cluster during ongoing distributed execution of a program, one or more nodes of the auxiliary computing node type may be selected for the removal.

Low stocks of coarse woody debris in a southwest Amazonian forest

Abstract: The stocks and dynamics of coarse woody debris (CWD) are significant components of the carbon cycle within tropical forests. However, to date, there have been no reports of CWD stocks and fluxes from the approximately 1.3 million km2 of lowland western Amazonian forests. Here, we present estimates of CWD stocks and annual CWD inputs from forests in southern Peru. Total stocks were low compared to other tropical forest sites, whether estimated by line-intercept sampling (24.4 ± 5.3 Mg ha−1) or by complete inventories within 11 permanent plots (17.7 ± 2.4 Mg ha−1). However, annual inputs, estimated from long-term data on tree mortality rates in the same plots, were similar to other studies (3.8 ± 0.2 or 2.9 ± 0.2 Mg ha−1 year−1, depending on the equation used to estimate biomass). Assuming the CWD pool is at steady state, the turnover time of coarse woody debris is low (4.7 ± 2.6 or 6.1 ± 2.6 years). These results indicate that these sites have not experienced a recent, large-scale disturbance event and emphasise the distinctive, rapid nature of carbon cycling in these western Amazonian forests.

Measuring Group Differences in High-Dimensional Choices: Method and Application to Congressional Speech

Abstract: We study the problem of measuring group differences in choices when the dimensionality of the choice set is large. We show that standard approaches suffer from a severe finite-sample bias, and we propose an estimator that applies recent advances in machine learning to address this bias. We apply this method to measure trends in the partisanship of congressional speech from 1873 to 2016, defining partisanship to be the ease with which an observer could infer a congressperson’s party from a single utterance. Our estimates imply that partisanship is far greater in recent years than in the past, and that it increased sharply in the early 1990s after remaining low and relatively constant over the preceding century.

Data storage integrity validation

Abstract: Embodiments of the present disclosure are directed to, among other things, validating the integrity of received and/or stored data payloads In some examples, a storage service may perform a first partitioning of a data object into first partitions based at least in part on a first operation The storage service may also verify the data object, by utilizing a verification algorithm, to generate a first verification value In some cases, the storage service may additionally perform a second partitioning of the data object into second partitions based at least in part on a second operation The second partitions may be different from the first partitions Additionally, the archival data storage service may verify the data object using the verification algorithm to generate a second verification value Further, the storage service may determine whether the second verification value equals the first verification value