计算机工程与应用
計算機工程與應用
계산궤공정여응용
COMPUTER ENGINEERING AND APPLICATIONS
2015年
16期
109-116
,共8页
戴春娥%陈维斌%傅顺开%李志强
戴春娥%陳維斌%傅順開%李誌彊
대춘아%진유빈%부순개%리지강
数据挖掘%GPU加速%并行计算%协同过滤
數據挖掘%GPU加速%併行計算%協同過濾
수거알굴%GPU가속%병행계산%협동과려
data mining%GPU accelerated%parallel computing%collaborative filtering
将计算密度高的部分迁移到GPU上是加速经典数据挖掘算法的有效途径。首先介绍GPU特性和主要的GPU编程模型，随后针对数据挖掘主要任务类型分别介绍基于GPU加速的工作，包括分类、聚类、关联分析、时序分析和深度学习。最后分别基于CPU和GPU实现协同过滤推荐的两类经典算法，并基于经典的MovieLens数据集的实验验证GPU对加速数据挖掘应用的显著效果，进一步了解GPU加速的工作原理和实际意义。
將計算密度高的部分遷移到GPU上是加速經典數據挖掘算法的有效途徑。首先介紹GPU特性和主要的GPU編程模型，隨後針對數據挖掘主要任務類型分彆介紹基于GPU加速的工作，包括分類、聚類、關聯分析、時序分析和深度學習。最後分彆基于CPU和GPU實現協同過濾推薦的兩類經典算法，併基于經典的MovieLens數據集的實驗驗證GPU對加速數據挖掘應用的顯著效果，進一步瞭解GPU加速的工作原理和實際意義。
장계산밀도고적부분천이도GPU상시가속경전수거알굴산법적유효도경。수선개소GPU특성화주요적GPU편정모형，수후침대수거알굴주요임무류형분별개소기우GPU가속적공작，포괄분류、취류、관련분석、시서분석화심도학습。최후분별기우CPU화GPU실현협동과려추천적량류경전산법，병기우경전적MovieLens수거집적실험험증GPU대가속수거알굴응용적현저효과，진일보료해GPU가속적공작원리화실제의의。
Transferring the procedure involving dense computation to GPU is known as an effective way to accelerate the whole procedure of many classical data mining algorithms. In this paper, features of GPU as well as existing programming models of GPU are introduced firstly. The representative works of fundamental data mining tasks are covered respectively, including classification, clustering, association analysis, time series analysis and deep learning. Two classical algorithms of collaborative filtering are implemented on CPU and GPU, and experiments with MovieLens data sets are conducted, which help to collect first-hand experience of applying GPU to accelerate the applications of data mining algorithms.
