加速基于 GPU 的 Top-K 计算.pdf

编号：29468

PDF 52页 1.42MB 下载积分：VIP专享

下载报告请您先登录！

加速基于 GPU 的 Top-K 计算.pdf

1、Christina Zhang efficient when k is small Disadvantage: depending on the implementation of PQ, the time complexity is usually at least proportional to log k, inefficient when k is large GPU implementation: WarpSelect from FAISS library (https:/ while PQ is commonly implemented with Heap for CPU, usi

2、ng parallel sorting can be more efficient on GPU. 7 ALGORITHM 3: SELECTION + FILTER a closely related problem is K-selection Problem: find the k-th smallest (or largest) element in a list of N elements (note some selection algorithms can be modified to return top-k results by carefully maintaining i

3、ntermediate data thus dont need the filter step) 8 ALGORITHM 3: SELECTION + FILTER many sorting algorithms have corresponding selection algorithms: advantages: selection algorithm usually has time complexity O(N), so efficient when k and N are large disadvantages: may need to read input multiple times GPU implementation: radix sort can be efficiently implemented on GPU, will explain radix select i

友情提示

1、下载报告失败解决办法
2、PDF文件下载后，可能会被浏览器默认打开，此种情况可以点击浏览器菜单，保存网页到桌面，就可以正常下载了。
3、本站不支持迅雷下载，请使用电脑自带的IE浏览器，或者360浏览器、谷歌浏览器下载即可。
4、本站报告下载后的文档和图纸-无水印,预览文档经过压缩，下载后原文更清晰。

本文（加速基于 GPU 的 Top-K 计算.pdf）为本站（X-iao）主动上传，三个皮匠报告文库仅提供信息存储空间，仅对用户上传内容的表现方式做保护处理，对上载内容本身不做任何修改或编辑。若此文所含内容侵犯了您的版权或隐私，请立即通知三个皮匠报告文库（点击联系客服），我们立即给予删除！

温馨提示：如果因为网速或其他原因下载失败请重新下载，重复下载不扣分。