Institution
Amazon.com
Company•Seattle, Washington, United States•
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.
Papers published on a yearly basis
Papers
More filters
•
29 Mar 2001TL;DR: In this article, a table is used to provide session-specific product recommendations to users that are based on the products viewed by the user during the current browsing session, and the table can also be used to supplement product detail pages with lists of related products.
Abstract: Various methods are disclosed for monitoring user browsing activities that indicate user interests in particular products or other items, and for using such information to identify items that are related to one another. In one embodiment, relationships between products within an online catalog are determined by identifying products that are frequently viewed by users within the same browsing session (e.g., products A and B are related because a significant portion of those who viewed A also viewed B). The resulting item relatedness data is preferably stored in a table that maps items to sets of related items. The table may be used to provide personalized product recommendations to users, and/or to supplement product detail pages with lists of related products. In one embodiment, the table is used to provide session-specific product recommendations to users that are based on the products viewed by the user during the current browsing session.
548 citations
••
06 Sep 2015TL;DR: It is shown empirically that the method can reduce the amount of communication by three orders of magnitude while training a typical DNN for acoustic modelling, and enables efficient scaling to more parallel GPU nodes than any other method that is aware of.
Abstract: We introduce a new method for scaling up distributed Stochastic Gradient Descent (SGD) training of Deep Neural Networks (DNN). The method solves the well-known communication bottleneck problem that arises for data-parallel SGD because compute nodes frequently need to synchronize a replica of the model. We solve it by purposefully controlling the rate of weight-update per individual weight, which is in contrast to the uniform update-rate customarily imposed by the size of a mini-batch. It is shown empirically that the method can reduce the amount of communication by three orders of magnitude while training a typical DNN for acoustic modelling. This reduction in communication bandwidth enables efficient scaling to more parallel GPU nodes than any other method that we are aware of, and it can be achieved with neither loss in convergence rate nor accuracy in the resulting DNN. Furthermore, the training can be performed on commodity cloud infrastructure and networking.
528 citations
••
TL;DR: In this paper, the authors show that tropical forest conversion, shifting cultivation and clearing of secondary vegetation make significant contributions to global emissions of greenhouse gases today, and have the potential for large additional emissions in future decades.
Abstract: Tropical forest conversion, shifting cultivation and clearing of secondary vegetation make significant contributions to global emissions of greenhouse gases today, and have the potential for large additional emissions in future decades. Globally, an estimated 3:1 10 9 t of biomass carbon of these types is exposed to burning annually, of which 1:1 10 9 t is emitted to the atmosphere through combustion and 49 10 6 t is converted to charcoal (including 26-31 10 6 t C of black carbon). The amount of biomass exposed to burning includes aboveground remains that failed to burn or decom- pose from clearing in previous years, and therefore exceeds the 1:9 10 9 t of aboveground biomass carbon cleared on average each year. Above- and belowground carbon emitted annually through decomposition processes totals 2:1 10 9 t C. A total gross emission (including decomposition of unburned aboveground biomass and of belowground biomass) of 3:41 10 9 t C year 1 results from clearing primary (nonfallow) and secondary (fallow) vegetation in the tropics. Adjustment for trace gas emissions using IPCC Second Assessment Report 100-year integration global warming potentials makes this equivalent to 3:39 10 9 to f CO 2-equivalent carbon under a low trace gas scenario and 3:8310 9 t under a high trace gas scenario. Of these totals, 1:0610 9 t (31%) is the result of biomass burning under the low trace gas scenario and 1:50 10 9 t (39%) under the high trace gas scenario. The net emissions from all clearing of natural vegetation and of secondary forests (including both biomass and soil fluxes) is 2:0 10 9 t C, equivalent to 2.0-2:4 10 9 to f CO 2-equivalent carbon. Adding emissions of 0:4 10 9 t C from land-use category changes other than deforestation brings the total for land-use change (not considering uptake of intact forest, recurrent burning of savannas or fires in intact forests) to 2 :4 10 9 t C, equivalent to 2.4-2:9 10 9 to f CO 2-equivalent carbon. The total net emission of carbon from the tropical land uses considered here (2:4 10 9 t C year 1 ) calculated for the 1981-1990 period is 50% higher than the 1:6 10 9 t C year 1 value used by the Intergovernmental Panel on Climate Change. The inferred (D 'missing') sink in the global carbon budget is larger than previously thought. However, about half of the additional source suggested here may be offset by a possible sink in uptake by Amazonian forests. Both alterations indicate that continued deforestation would produce greater impact on global carbon emissions. The total net emission of carbon calculated here indicates a major global warming impact from tropical land uses, equivalent to approximately 29% of the total anthropogenic emission from fossil fuels and land-use change.
523 citations
•
17 May 1996
TL;DR: In this article, a method and system for securely indicating to a customer one or more credit card numbers that a merchant has on file for the customer when communicating with the customer over a non-secure network is presented.
Abstract: A method and system for securely indicating to a customer one or more credit card numbers that a merchant has on file for the customer when communicating with the customer over a non-secure network. The merchant sends a message to the customer that contains only a portion of each of the credit card numbers that are on file with the merchant. The message may also contain a notation explaining which portion of each of the credit card numbers has been extracted. A computer (38) retrieves the credit card numbers on file for the customer in a database (40), constructs the message, and transmits the message to a customer location (10) over the Internet network (30), or other non-secure network. The customer can then confirm in a return message that a specific one of the credit card numbers on file with the merchant should be used in charging a transaction. Since only a portion of the credit card number(s) are included in any message transmitted, a third party cannot discover the customer's complete credit card number(s).
522 citations
••
10 Apr 2016TL;DR: This work formulate bitrate adaptation as a utility maximization problem and devise an online control algorithm called BOLA that uses Lyapunov optimization techniques to minimize rebuffering and maximize video quality and proves that B OLA achieves a time-average utility that is within an additive term O(1/V) of the optimal value.
Abstract: Modern video players employ complex algorithms to adapt the bitrate of the video that is shown to the user. Bitrate adaptation requires a tradeoff between reducing the probability that the video freezes and enhancing the quality of the video shown to the user. A bitrate that is too high leads to frequent video freezes (i.e., rebuffering), while a bitrate that is too low leads to poor video quality. Video providers segment the video into short chunks and encode each chunk at multiple bitrates. The video player adaptively chooses the bitrate of each chunk that is downloaded, possibly choosing different bitrates for successive chunks. While bitrate adaptation holds the key to a good quality of experience for the user, current video players use ad-hoc algorithms that are poorly understood. We formulate bitrate adaptation as a utility maximization problem and devise an online control algorithm called BOLA that uses Lyapunov optimization techniques to minimize rebuffering and maximize video quality. We prove that BOLA achieves a time-average utility that is within an additive term O(1/V) of the optimal value, for a control parameter V related to the video buffer size. Further, unlike prior work, our algorithm does not require any prediction of available network bandwidth. We empirically validate our algorithm in a simulated network environment using an extensive collection of network traces. We show that our algorithm achieves near-optimal utility and in many cases significantly higher utility than current state-of-the-art algorithms. Our work has immediate impact on real-world video players and BOLA is part of the reference player implementation for the evolving DASH standard for video transmission.
508 citations
Authors
Showing all 13498 results
Name | H-index | Papers | Citations |
---|---|---|---|
Jiawei Han | 168 | 1233 | 143427 |
Bernhard Schölkopf | 148 | 1092 | 149492 |
Christos Faloutsos | 127 | 789 | 77746 |
Alexander J. Smola | 122 | 434 | 110222 |
Rama Chellappa | 120 | 1031 | 62865 |
William F. Laurance | 118 | 470 | 56464 |
Andrew McCallum | 113 | 472 | 78240 |
Michael J. Black | 112 | 429 | 51810 |
David Heckerman | 109 | 483 | 62668 |
Larry S. Davis | 107 | 693 | 49714 |
Chris M. Wood | 102 | 795 | 43076 |
Pietro Perona | 102 | 414 | 94870 |
Guido W. Imbens | 97 | 352 | 64430 |
W. Bruce Croft | 97 | 426 | 39918 |
Chunhua Shen | 93 | 681 | 37468 |