Evaluating Fast Algorithms for Convolutional Neural Networks on

第十届中国R会议【直播】

233018 人浏览

已收藏收藏

8488

微信分享

视频列表

圆桌讨论

中产阶级如何利用量化投资工具完成财富进阶

统计转移学习（及其在统计过程控制的应用）

大数据时代下的统计学思维—以文本挖掘为例

Building Better Connected World with Artificial Intelligence Tec

李航

5142

Statistical learning with genomic big-data

刘军

5069

Data Science, machine learning, precision medicine, and all that

王永雄

5852

Adaptive False Discovery Rate regression with application in int

条件随机场及其在生物信息学中的应用

Glimma: getting greater graphics for your genes

Charity Law

6076

Rsubread: an efficient toolkit for mapping and counting short se

汽车消费的数字化决策

机器学习在滴滴

摩拜单车的数据科学实践

A smooth simultaneous confidence band for correlation curve

张园园

4811

Simultaneous conficence bands for mean and variance function bas

蔡利

4091

Prediction Interval for Autoregressive Time Series via Oracally

顾莉洁

5109

On a vector double autoregressive model

张兴发

4107

Banded Spatio-Temporal Autoregressions with with Application to

基于车辆GPS 数据的交通大数据应用

Branding with social media: User gratifications, usage patterns,

质保数据建模与分析

卓越质量管理中的大数据分析

PM 2.5 数据的时空特征及统计建模

不可或缺的优质地理大数据

基于室内定位数据（IPS）的时空行为分析

地理数据与商业网点选址实战

Identification of disease-causing single nucleotide variants in

癌症转录组大数据的可视化与再挖掘

消化道肿瘤基因组学研究进展

临床医生眼中的医疗大数据研究：需求和挑战

R Usage in Pharmaceutical Industry

手机数据与经济活动测度

词汇、概念、数字：文本探勘技术于中国近代观念史研究中的应用与实践

邱伟云

4767

群像的描绘与类型的分析：用数字工具挖掘《德意志人物志》

王涛

5109

network diffusion: Simulate and Visualize Network Diffusion

计算与人文：作为新领域的“数字人文”

Elastic Stack 与机器学习

Evaluating Fast Algorithms for Convolutional Neural Networks on

卢丽强

8489

Pluto: A Distributed Heterogeneous Deep Learning Framework

王思宇

7009

Exploring Heterogeneous Algorithms for Accelerating Deep Convolu

嵌入式上的深度学习初探

再抽样法分析夫妻般配与家庭工资不平等

法律的定量分析及其实践

Latent Variable Modeling for Cognitive Assessment Through Second

刘京辰

5802

中文文本分析方便工具包chinese.misc介绍

大数据教育平台的建设与探索

大数据教育学科建设的关键因素

微启的旋转门：大数据教育界与工业界的生态进化

赵鹏

4350

跟踪 R 社区动态 - R Weekly 的背后

覃文锋

4418

Learning R Internals and C++ via Rcpp

任乾

4319

Persistent Reproducible Reporting with Docker and R

肖楠

4709

利用css对shiny页面优化及利用htmlwidgets包创建HTML控件

SAS统计图表：一键式的图表生成术

Interaction+: “让可视化动起来”的既有网页交互

数控机床大数据分析

人工智能颠覆客服行业的实践

机器学习在营销管理中的应用

从统计学生到互金数据科学家之路

“AI+慢性病管理”使精准医疗成为可能

心理学在助老机器人研发中的应用

心理学研究规范化及在R语言的实现

R语言在加强心理学可重复性中的作用

基于R与Rstudio的心理统计教学模式探索

基于社会媒体大数据的心理学研究

NLP在金融报告自动化的实践

面向社交媒体的商业大数据挖掘

自然语言处理在医疗智能辅助中的应用

智能时代的量化资产管理

从语言智能到法务智能

bandit算法与推荐系统

腾讯社交广告实践中智能出价新模式：oCPA

王流斌

6647

Learning theory for deep nets

林绍波

5156

Triple Generative Adversarial Networks

Kaggle 数据挖掘比赛经验分享

On equivalence of likelihood maximization of stochastic block mo

张忠元

4286

从文本分析看小说中人物的复杂关系：以琅琊榜为例

周静

6554

R语言中的深度学习: 用Mxnet进行车型识别

条件GAN用于车型设计和判别

增强学习打麻将

视频介绍

课程资料

评价

嘉宾介绍

主题介绍

In recent years, Convolutional Neural Networks (CNNs) have become widely adopted for computer vision tasks. FPGAs have been adequately explored as a promising hardware accelerator for CNNs due to its high performance, energy efficiency, and reconfigurability. However, prior FPGA solutions based on the conventional convolutional algorithm is often bounded by the computational capability of FPGAs (e.g., the number of DSPs). In this paper, we demonstrate that fast Winograd algorithm can dramatically reduce the arithmetic complexity, and improve the performance of CNNs on FPGAs. We first propose a novel architecture for implementing Winograd algorithm on FPGAs. Our design employs line buffer to effectively reuse the feature map data among different tiles. We also effectively pipeline the Winograd PE engine and initiate multiple PEs through parallelization. Meanwhile, there exists a complex design space to explore. We propose an analytical model to predict the resource usage and reason about the performance. Then, we use the model to guide a fast design space exploration. Experiments using the state-of-the-art CNNs demonstrate the best performance and energy efficiency on FPGAs. We achieve an average 785.1 GOP/s for the convolutional layers and 749.4 GOP/s for the overall AlexNet and an average 2653.4 GOP/s for the convolutional layers and 2272.6 GOP/s for the overall VGG16 on Xilinx ZCU102 platform.

未上传任何附件

说点什么

—— 点击加载更多 ——

收起

为你推荐

啊哦，暂无相关推荐