Institution
Samsung
Company•Seoul, South Korea•
About: Samsung is a company organization based out in Seoul, South Korea. It is known for research contribution in the topics: Layer (electronics) & Signal. The organization has 134067 authors who have published 163691 publications receiving 2057505 citations. The organization is also known as: Samsung Group & Samsung chaebol.
Papers published on a yearly basis
Papers
More filters
••
15 Jun 2019TL;DR: In this article, a quantization interval learning (QIL) method is proposed to quantize activations and weights via a trainable quantizer that transforms and discretizes them.
Abstract: Reducing bit-widths of activations and weights of deep networks makes it efficient to compute and store them in memory, which is crucial in their deployments to resource-limited devices, such as mobile phones. However, decreasing bit-widths with quantization generally yields drastically degraded accuracy. To tackle this problem, we propose to learn to quantize activations and weights via a trainable quantizer that transforms and discretizes them. Specifically, we parameterize the quantization intervals and obtain their optimal values by directly minimizing the task loss of the network. This quantization-interval-learning (QIL) allows the quantized networks to maintain the accuracy of the full-precision (32-bit) networks with bit-width as low as 4-bit and minimize the accuracy degeneration with further bit-width reduction (i.e., 3 and 2-bit). Moreover, our quantizer can be trained on a heterogeneous dataset, and thus can be used to quantize pretrained networks without access to their training data. We demonstrate the effectiveness of our trainable quantizer on ImageNet dataset with various network architectures such as ResNet-18, -34 and AlexNet, on which it outperforms existing methods to achieve the state-of-the-art accuracy.
269 citations
•
29 Dec 2000TL;DR: In this paper, the authors have disclosed control circuitry capable of being selectively set to disable the transmission of information concerning the location of the wireless mobile station, where the control circuitry also comprises a directory of telephone numbers of locations authorized to receive information about the locations of the mobile station.
Abstract: In a wireless mobile station of the type having a position locating system capable of determining the location of the wireless mobile station, there is disclosed control circuitry capable of being selectively set to disable the transmission of information concerning the location of the wireless mobile station. The control circuitry also comprises a directory of telephone numbers of locations authorized to receive information concerning the location of the wireless mobile station. Also disclosed is control circuitry capable of receiving a code that causes the wireless mobile station to transmit information concerning the location of the wireless mobile station. Also disclosed are methods for selectively disabling the transmission of information concerning the location of the wireless mobile station.
269 citations
••
19 Mar 2018TL;DR: This work comprehensively analyzes the energy and performance impact of data movement for several widely-used Google consumer workloads, and finds that processing-in-memory (PIM) can significantly reduceData movement for all of these workloads by performing part of the computation close to memory.
Abstract: We are experiencing an explosive growth in the number of consumer devices, including smartphones, tablets, web-based computers such as Chromebooks, and wearable devices. For this class of devices, energy efficiency is a first-class concern due to the limited battery capacity and thermal power budget. We find that data movement is a major contributor to the total system energy and execution time in consumer devices. The energy and performance costs of moving data between the memory system and the compute units are significantly higher than the costs of computation. As a result, addressing data movement is crucial for consumer devices. In this work, we comprehensively analyze the energy and performance impact of data movement for several widely-used Google consumer workloads: (1) the Chrome web browser; (2) TensorFlow Mobile, Google's machine learning framework; (3) video playback, and (4) video capture, both of which are used in many video services such as YouTube and Google Hangouts. We find that processing-in-memory (PIM) can significantly reduce data movement for all of these workloads, by performing part of the computation close to memory. Each workload contains simple primitives and functions that contribute to a significant amount of the overall data movement. We investigate whether these primitives and functions are feasible to implement using PIM, given the limited area and power constraints of consumer devices. Our analysis shows that offloading these primitives to PIM logic, consisting of either simple cores or specialized accelerators, eliminates a large amount of data movement, and significantly reduces total system energy (by an average of 55.4% across the workloads) and execution time (by an average of 54.2%).
267 citations
••
13 Dec 2004TL;DR: In this article, a high-density 64Mbit PRAM was successfully fabricated using N-doped Ge/sub 2/Sb/Sub 2/Te/sub 5/ (GST) and optimal GST etching process, achieving low writing current of 0.6 mA and clear separation between SET and RESET resistance distributions.
Abstract: Highly manufacturable 64Mbit PRAM has been successfully fabricated using N-doped Ge/sub 2/Sb/sub 2/Te/sub 5/ (GST) and optimal GST etching process. Using those technologies, it was possible to achieve the low writing current of 0.6 mA and clear separation between SET and RESET resistance distributions. The 64Mb PRAM was designed to support commercial NOR flash memory compatible interfaces. Therefore, the fabricated chip was tested under the mobile application platform and its functionality and reliability has been evaluated by operation temperature dependency, disturbance, endurance, and retention. Finally, it was clearly demonstrated that high density PRAM can be fabricated in the product level with strong reliability to produce new nonvolatile memory markets.
267 citations
••
TL;DR: In this paper, a three-phase heuristic for the problem of minimizing the total weighted tardiness on a single machine in the presence of sequence-dependent setup times is proposed.
Abstract: We propose a three-phase heuristic for the problem of minimizing the total weighted tardiness on a single machine in the presence of sequence-dependent setup times. In the first phase a number of parameters characterizing the problem instance at hand are calculated. In the second phase we develop a schedule by using a new priority rule whose parameters are calculated based on the results of the first phase. Computational experiments show that this rule significantly outperforms the only other rule so far developed in the literature. The third phase consists of a local improvement procedure to improve the schedule obtained in the second phase. The procedure we suggest has been successfully implemented in an industrial scheduling system.
267 citations
Authors
Showing all 134111 results
Name | H-index | Papers | Citations |
---|---|---|---|
Yi Cui | 220 | 1015 | 199725 |
Hyun-Chul Kim | 176 | 4076 | 183227 |
Hannes Jung | 159 | 2069 | 125069 |
Yongsun Kim | 156 | 2588 | 145619 |
Yu Huang | 136 | 1492 | 89209 |
Robert W. Heath | 128 | 1049 | 73171 |
Shuicheng Yan | 123 | 810 | 66192 |
Shi Xue Dou | 122 | 2028 | 74031 |
Young Hee Lee | 122 | 1168 | 61107 |
Alan L. Yuille | 119 | 804 | 78054 |
Yang-Kook Sun | 117 | 781 | 58912 |
Sang Yup Lee | 117 | 1005 | 53257 |
Guoxiu Wang | 117 | 654 | 46145 |
Richard G. Baraniuk | 107 | 770 | 57550 |
Jef D. Boeke | 106 | 456 | 52598 |