scispace - formally typeset
Search or ask a question
Institution

NEC

CompanyTokyo, Japan
About: NEC is a company organization based out in Tokyo, Japan. It is known for research contribution in the topics: Signal & Layer (electronics). The organization has 33269 authors who have published 57670 publications receiving 835952 citations. The organization is also known as: NEC Corporation & NEC Electronics Corporation.


Papers
More filters
Proceedings ArticleDOI
03 Sep 2010
TL;DR: It is proposed to build and deploy an open - but backward compatible - wireless network infrastructure that can be easily deployed on college campuses worldwide, and allows rapid innovation of network services, contributed by researchers, network operators, equipment vendors and third party developers.
Abstract: In the past couple of years we've seen quite a change in the wireless industry: Handsets have become mobile computers running user-contributed applications on (potentially) open operating systems. It seems we are on a path towards a more open ecosystem; one that has been previously closed and proprietary. The biggest winners are the users, who will have more choice among competing, innovative ideas.The same cannot be said for the wireless network infrastructure, which remains closed and (mostly) proprietary, and where innovation is bogged down by a glacial standards process. Yet as users, we are surrounded by abundant wireless capacity and multiple wireless networks (WiFi and cellular), with most of the capacity off-limits to us. It seems industry has little incentive to change, preferring to hold onto control as long as possible, keeping an inefficient and closed system in place.This paper is a "call to arms" to the research community to help move the network forward on a path to greater openness. We envision a world in which users can move freely between any wireless infrastructure, while providing payment to infrastructure owners, encouraging continued investment. We think the best path to get there is to separate the network service from the underlying physical infrastructure, and allow rapid innovation of network services, contributed by researchers, network operators, equipment vendors and third party developers.We propose to build and deploy an open - but backward compatible - wireless network infrastructure that can be easily deployed on college campuses worldwide. Through virtualization, we allow researchers to experiment with new network services directly in their production network.

208 citations

Patent
23 Dec 2002
TL;DR: In this article, the RS-CVD system was used to improve the film quality of a silicon oxide film by adding nitrogen atom containing gas (N 2 gas), NO gas, N 2 O gas, NO 2 gas or the like) to oxygen atom containing gases (O 2 gas, O 3 gas, etc.) to increase the quantity of the atomic oxygen generated by the plasmas in the plasma generating space.
Abstract: An object of this invention is to provide a silicon oxide film formation method capable of enhancing efficiency for generating atomic oxygen and improving film quality of a silicon film (SiO 2 film) in forming the silicon oxide film using an RS-CVD system. This object is attained by adding nitrogen atom containing gas (N 2 gas, NO gas, N 2 O gas, NO 2 gas or the like) to oxygen atom containing gas (O 2 gas, O 3 gas or the like) introduced into a plasma generating space in the vacuum container to produce plasmas by these gases and to thereby increase the quantity of the atomic oxygen generated by the plasmas in the plasma generating space.

208 citations

Posted Content
Felix Wu1, Angela Fan2, Alexei Baevski2, Yann N. Dauphin3, Michael Auli2 
TL;DR: It is shown that a very lightweight convolution can perform competitively to the best reported self-attention results, and dynamic convolutions are introduced which are simpler and more efficient than self-ATTention.
Abstract: Self-attention is a useful mechanism to build generative models for language and images. It determines the importance of context elements by comparing each element to the current time step. In this paper, we show that a very lightweight convolution can perform competitively to the best reported self-attention results. Next, we introduce dynamic convolutions which are simpler and more efficient than self-attention. We predict separate convolution kernels based solely on the current time-step in order to determine the importance of context elements. The number of operations required by this approach scales linearly in the input length, whereas self-attention is quadratic. Experiments on large-scale machine translation, language modeling and abstractive summarization show that dynamic convolutions improve over strong self-attention models. On the WMT'14 English-German test set dynamic convolutions achieve a new state of the art of 29.7 BLEU.

208 citations

Patent
Kazuya Suzuki1, Masahiro Jibiki1
10 Jun 2004
TL;DR: In this paper, the route information used for dynamic route control can be shared by an active router and a standby router even in normal operation, and the route calculation is performed according to a routing protocol, enabling efficient operation.
Abstract: Route information used for dynamic route control can be shared by an active router and a standby router even in normal operation. After route calculation is performed according to a routing protocol, the calculated route information can be shared, enabling efficient operation. Protocol engine sections (31a to 31d) communicate with routers (41, 42), calculate a route to be selected for the routing according to the corresponding protocol. When the route information is altered, a route sharing information transmitting section (33) transmits the route information to the standby router. The protocol engine section in each router acting as an active router communicates with the router (41, 42), calculates a route to be selected for the routing according to the corresponding protocol, and transmits the route information, if it is altered. A route sharing information receiving section (34) receives the route information.

207 citations

Journal ArticleDOI
TL;DR: In this paper, the pseudopotential total energy of graphitic carbon nitride is investigated by means of first principles pseudoprocessor total energy calculation, and the stability of single layers of C3N4 and CN is examined by geometrical optimization and phonon frequencies.

206 citations


Authors

Showing all 33297 results

NameH-indexPapersCitations
Pulickel M. Ajayan1761223136241
Xiaodong Wang1351573117552
S. Shankar Sastry12285886155
Sumio Iijima106633101834
Thomas W. Ebbesen9930570789
Kishor S. Trivedi9569836816
Sharad Malik9561537258
Shigeo Ohno9130328104
Adrian Perrig8937453367
Jan M. Rabaey8152536523
C. Lee Giles8053625636
Edward A. Lee7846234620
Otto Zhou7432218968
Katsumi Kaneko7458128619
Guido Groeseneken73107426977
Network Information
Related Institutions (5)
Bell Labs
59.8K papers, 3.1M citations

92% related

Hitachi
101.4K papers, 1.4M citations

92% related

Samsung
163.6K papers, 2M citations

91% related

IBM
253.9K papers, 7.4M citations

91% related

Intel
68.8K papers, 1.6M citations

90% related

Performance
Metrics
No. of papers from the Institution in previous years
YearPapers
20238
202220
2021234
2020518
2019952
20181,088