ai-bigdata.html 56 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326
  1. <!DOCTYPE html>
  2. <html>
  3. <head>
  4. <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
  5. <title>AI、大数据定制开发-铭讯科技</title>
  6. <!-- Stylesheets -->
  7. <link href="static/css/bootstrap.css" rel="stylesheet">
  8. <link href="static/css/style.css" rel="stylesheet">
  9. <link href="static/css/slick.css" rel="stylesheet">
  10. <link href="static/css/responsive.css" rel="stylesheet">
  11. <!--Color Switcher Mockup-->
  12. <link href="static/css/color-switcher-design.css" rel="stylesheet">
  13. <!--Color Themes-->
  14. <link id="theme-color-file" href="static/css/official-blue-theme.css" rel="stylesheet">
  15. <link rel="icon" type="image/x-icon" href="./static/image/icon.jpg"/>
  16. <link href="static/css/30.css" rel="stylesheet"/>
  17. <!-- Responsive -->
  18. <meta http-equiv="X-UA-Compatible" content="IE=edge">
  19. <meta name="viewport" content="width=device-width, initial-scale=1.0, maximum-scale=1.0, user-scalable=0">
  20. <!--[if lt IE 9]>
  21. <script src="/officialBlue/js/html5shiv.js"></script><![endif]-->
  22. <!--[if lt IE 9]>
  23. <script src="/officialBlue/js/respond.js"></script><![endif]-->
  24. <style>
  25. @media screen and (max-width:480px){
  26. .sidebar .sidebar-blog-category{
  27. display: none;
  28. }
  29. .footer .footer-block .footer-tit{
  30. width: 100px;
  31. margin: 0 auto 5px auto;
  32. }
  33. .footer .footer-block .hot-line{
  34. width: 340px;
  35. margin: 0;
  36. margin-left: -120px;
  37. }
  38. .footer .footer-block p>span{
  39. display:none;
  40. }
  41. }
  42. </style>
  43. </head>
  44. <body>
  45. <div class="page-wrapper">
  46. <!-- Main Header-->
  47. <!--顶部菜单 -->
  48. <div class="header-perch">
  49. <div class="hx-header-fix">
  50. <div class="auto-container">
  51. <div class="hx-header mobile">
  52. <div class="hx-logo" style="width: 108px;height: 70px;">
  53. <a href="index1.html">
  54. <img style="width: 100%;" src="static/image/login.png" class="logo_blue" alt="">
  55. </a>
  56. </div>
  57. <div class="menu-operate">
  58. <div id="menu-operate-close"><img src="static/picture/71.png" alt=""></div>
  59. <div id="menu-operate-spread"><img src="static/picture/72.png" alt=""></div>
  60. </div>
  61. <div class="hx-menu">
  62. <ul>
  63. <li><a href="index1.html">首页</a><span class="btm-line"><img src="static/picture/selected.png" alt=""></span></li>
  64. <li><a href="services.html">软件开发服务</a><span class="btm-line"><img src="static/picture/selected.png" alt=""></span></li>
  65. <li><a href="case.html">服务案例</a><span class="btm-line"><img src="static/picture/selected.png" alt=""></span></li>
  66. <li><a href="faq.html">新闻资讯</a><span class="btm-line"><img src="static/picture/selected.png" alt=""></span></li>
  67. <li><a href="contact.html">联系我们</a><span class="btm-line"><img src="static/picture/selected.png" alt=""></span></li>
  68. <li><a href="about.html">关于我们</a><span class="btm-line"><img src="static/picture/selected.png" alt=""></span></li>
  69. <li style="width: 180px;"><a href=""> 咨询电话:17347315117 </a><span class="btm-line"><img src="static/picture/selected.png" alt=""></span></li>
  70. </ul>
  71. </div>
  72. </div>
  73. </div>
  74. </div>
  75. </div>
  76. <!--End Main Header -->
  77. <!--Page Title-->
  78. <section class="page-title" style="background-image:url(static/picture/w-R.png);">
  79. <div class="auto-container">
  80. <h2>AI、大数据开发</h2>
  81. </div>
  82. </section>
  83. <!--Breadcrumb-->
  84. <div class="Dg" style=" width: 1190px;margin:0 auto;; margin-top: 10px;"> <a style="margin-left: 15px;" href='index1.html' class=''>主页</a> > <a href='services.html'>软件开发服务</a> > <a href=""> AI,大数据开发</a></div>
  85. <!--End Page Title-->
  86. <!--Sidebar Page Container-->
  87. <div class="sidebar-page-container">
  88. <div class="auto-container">
  89. <div class="row clearfix">
  90. <div class="content-side col-lg-8 col-md-12 col-sm-12">
  91. <div class="services-single">
  92. <div class="inner-service">
  93. <div class="lower-content">
  94. <h2 style="text-align: center">人工智能、大数据开发介绍</h2>
  95. <div class="text">
  96. <p><span style="color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;">人工智能是近年来科技发展的重要方向,在大数据时代,对数据采集、挖掘、应用的技术越来越受到瞩目。在人工智能和大数据的开发过程中,有哪些特别需要注意的要点?</span><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><span style="color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;">  人工智能领域的算法大师、华盛顿大学教授Pedro Domingos对此进行了深入思考。</span><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><span style="color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;">  在我们新近翻译的《智能Web算法》(第2版)中,对Pedro Domingos教授的观点进行了高度的概括,提炼出12个注意点,为行业开发实践提供了重要参考:</span><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><span style="box-sizing: border-box; outline: 0px; font-weight: bold; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;">  注意点1:你的数据未必可靠</span><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><span style="color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;">  在实际应用中,有很多各种各样的原因会导致你的数据是不可靠的。因此,当你将数据用于解决问题前,必须经常留心来检查数据是否值得信赖。如果基于糟糕的数据来挖掘,无论多么聪明的人也永远只会获得糟糕的结果。下面列举了一些常见的可导致数据可靠性问题的因素:</span><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><span style="color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;">  用于开发的数据,往往和实际情况下的数据分布不同。例如也许你想把用户按照身高划分为&ldquo;高&rdquo;、&ldquo;中等&rdquo;、&ldquo;矮&rdquo;三档,但如果系统开发时使用的数据集里最低用户的身高是6英尺(184cm),那么很有可能你开发出来的系统里会把一个&ldquo;仅有6英尺&rdquo;的用户称为&ldquo;矮&rdquo;用户</span><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><span style="color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;">  你的数据集中存在很多缺失数据。事实上,除非是人为构造的数据集合,否则很难避免缺失数据问题的发生,如何处理数据缺失的问题是很有技巧的事情。实践中我们要么是干脆丢弃一部分残缺的数据,要么就是想办法计算一些数值去填补这些缺失值。无论哪种方法都可能导致应用结果的不稳定</span><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><span style="color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;">  你的数据可能随时在变化。数据库的表结构可能会变,数据定义也可能会变</span><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><span style="color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;">  你的数据可能没有被归一化。假设你可能在观察一组用户的体重,为了能够获得有效的结论,首先需要对每个体重的衡量单位进行归一化,是英镑还是公斤,不能混淆着用</span><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><span style="color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;">  你的数据可能并不适用于相应的算法。数据存在着各种各样的形式和规范,或者叫数据类型(data types),有些是数值化的数据,有些则不是。有些数据集合能被有序排列,有些则做不到。有些是离散化的数据(例如房间里的人数),另一些则是连续化的(例如气温或者气压等数据)</span><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><span style="box-sizing: border-box; outline: 0px; font-weight: bold; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;">  注意点2:计算难以瞬间完成</span><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><span style="color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;">  完成任何一个人工智能解决方案的计算,都需要一定的时间,方案的响应速度,对商业应用的成功与否起到十分关键的作用。不能总是盲目假设任何算法在所有数据集上都一定能在规定时间内完成,你需要测试下算法的性能是否在可接受的应用范围内。</span><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><span style="color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;">  以搜索引擎为例,用户对结果返回的时长是有忍耐的限度的。如果用户等待的时间超过10秒,50%的用户会流失,如果等待时间超过1分钟,90%以上的用户会流失。在开发智能应用系统时,不能为了达到更好的算法精度而忽略系统运算和等待的时间,否则会导致整个产品的失败。</span><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><span style="box-sizing: border-box; outline: 0px; font-weight: bold; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;">  注意点3:数据的规模非常重要</span><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><span style="color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;">  当我们考虑智能应用时,数据规模是很重要的因素。数据规模的影响可以分为两点来考察:第一点是规模会影响应用系统的响应速度,上一节我们刚提过;第二点是在很大的数据集上的挖掘出有价值结果的能力会受到考验。例如为100个用户开发的电影或音乐推荐系统可能效果很好,但是同样的算法移植到有着100000个用户的环境里,效果可能就不尽如人意了。</span><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><span style="color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;">  其次,使用更多的数据来训练的简单算法,比受制于维度诅咒(Dimension Curse)的复杂算法往往有好得多的效果。类似Google这样拥有海量数据的大型企业,优秀的应用效果不仅来自于精妙复杂的算法,也来自于其对海量训练数据的大规模分析挖掘。(达观数据 陈运文)</span><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><span style="box-sizing: border-box; outline: 0px; font-weight: bold; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;">  注意点4:不同的算法具有不同的扩展能力</span><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><span style="color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;">  我们不能假设智能应用系统都可以通过简单增加服务器的方法来扩展性能。有些算法是有扩展性的,而另一些则不行。</span><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><span style="color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;">  例如如果我们要从数亿的文章标题里,找出标题相似的各个组的文章,注意并不是所有的聚类算法此时都能并行化运行的,你应该在设计系统的同时就考虑可扩展性。有些情况下你需要将数据切分成较小的集合,并能够让智能算法在各个集合上并行运行。设计系统时所选择的算法,往往需要有并行化的版本,而在一开始就需要将其纳入考虑,因为通常围绕着算法还会有很多相关联的商业逻辑和体系结构需要一并考虑。</span><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><span style="box-sizing: border-box; outline: 0px; font-weight: bold; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;">  注意点5:并不存在万能的方法</span><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><span style="color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;">  你可能听说过一句谚语&ldquo;当你有了把榔头的时候,看什么东西都像钉子&rdquo;,这里想表达的意思是:并不存在能够解决所有智能应用问题的万能算法。</span><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><span style="color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;">  智能应用软件和其他所有软件类似&mdash;&mdash;具有其特定的应用领域和局限性。当面对新的应用领域时,一定要充分的验证原有方法的可行性,而且你最好能尝试用全新的视角来考察问题,因为不同的算法在解决特定的问题时才会更有效和得当(达观数据 陈运文)。</span><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><span style="box-sizing: border-box; outline: 0px; font-weight: bold; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;">  注意点6:数据并不是万能的</span><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><span style="color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;">  根本上看,机器学习算法并不是魔法,它需要从训练数据开始,逐步延伸到未知数据中去。</span><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><span style="color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;">  例如假设你已经对数据的分布规律有所了解,那么通过图模型来表达这些先验的知识会非常有效。除了数据以外,你还需要仔细的考虑,该领域有哪些先验知识可以应用,这对开发一个更有效的分类器会很有帮助。数据和行业经验结合往往能事半功倍。</span><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><span style="box-sizing: border-box; outline: 0px; font-weight: bold; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;">  注意点7:模型训练的时间差异很大</span><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><span style="color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;">  在特定应用中,可能某些参数的微小变化就会让模型的训练时间出现很大的差异。例如在深度神经网络训练时就会有各种各样的参数调节的情况发生。</span><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><span style="color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;">  人们往往会直观地觉得调整参数时,训练时间是基本稳定不变的。例如假设有个系统是计算地球平面上任意两点之间的距离的,那么任意给出两个点的坐标时,计算时间差不多都是相同的。但在另一些系统里却并非如此,有时细微的调整会带来很明显的时间差异,有时差异甚至可以大到数小时,而不是数秒。</span><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><span style="box-sizing: border-box; outline: 0px; font-weight: bold; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;">  注意点8:泛化能力是目标</span><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><span style="color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;">  机器学习实践中最普遍存在的一个误区是陷入处理细节中而忘了最初的目标&mdash;&mdash;通过调查来获得处理问题的普适的方法。</span><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><span style="color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;">  测试阶段是验证某个方法是否具备泛化能力(generalization ability)的关键环节(通过交叉验证、外部数据验证等方法),但是寻找合适的验证数据集不容易。如果在一个只有几百个样本的集合上去训练有数百万维特征的模型,试图想获得优秀的精度是很荒唐的。</span><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><span style="box-sizing: border-box; outline: 0px; font-weight: bold; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;">  注意点9:人类的直觉未必准确</span><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><span style="color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;">  在特征空间膨胀的时候,输入信息间形成的组合关系会快速增加,这让人很难像对中等数据集合那样能够对其中一部分数据进行抽样观察。更麻烦的是,特征数量增加时人类对数据的直觉会迅速降低。</span><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><span style="color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;">  例如在高维空间里,多元高斯分布并不是沿着均值分布,而是像一个扇贝形状围绕在均值附近,这和人们的主观感受完全不同。在低维空间中建立一个分类器并不难,但是当维度增加时,人类就很难直观的理解了。</span><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><span style="box-sizing: border-box; outline: 0px; font-weight: bold; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;">  注意点10:要考虑融入更多新特征</span><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><span style="color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;">  你很可能听说过谚语&ldquo;进来的是垃圾,出去的也是垃圾&rdquo;(garbage in, garbage out),在建立机器学习应用中这一点尤其重要。为了避免挖掘的效果失控,关键是要充分掌握问题所在的领域,通过调查数据来生成各种各样的特征,这样的做法会对提升分类的准确率和泛化能力有很大的帮助。仅靠把数据扔进分类器就想获得优秀结果的幻想是不可能实现的。</span><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><span style="box-sizing: border-box; outline: 0px; font-weight: bold; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;">  注意点11:要学习各种不同的模型</span><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><span style="color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;">  模型的组合(Ensemble)技术正变得越来越流行了,因为组合方法,仅需要付出少许偏见(bias)的代价,就能大大的减少算法的不确定性。在着名的Netflix算法竞赛中,冠军队以及成绩优异队伍们全都使用了组合模型方法,把超过100个模型合并在一起(在模型上叠加高层的模型形成组合)以提升效果。在人工智能用于实际应用时,从业者普遍都认为,未来的算法一定时会通过模型组合的方法来获得更好精度,但是这也会抬高非专业人员理解系统机制的门槛。</span><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><span style="box-sizing: border-box; outline: 0px; font-weight: bold; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;">  注意点12:相关关系不等同于因果关系</span><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><span style="color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;">  这一点值得反复强调,我们可以通过一句调侃的话来解释:&ldquo;地球变暖、地震、龙卷风,以及其他自然灾害,都和18世纪以来全球海盗数量的减少有直接关系&rdquo;。这两个变量的变化有相关性,但是并不能说存在因果关系,因为往往存在第三类(甚至第4、5类)未被观察到的变量在起作用。相关关系应该看作是潜在的因果关系的一定程度的体现,但需要进一步研究。</span><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><br style="box-sizing: border-box; outline: 0px; overflow-wrap: break-word; color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;"><span style="color: #4d4d4d; font-family: -apple-system, 'SF UI Text', Arial, 'PingFang SC', 'Hiragino Sans GB', 'Microsoft YaHei', 'WenQuanYi Micro Hei', sans-serif, SimHei, SimSun; font-size: 16px; background-color: #ffffff;">  在开发人工智能与大数据应用系统时,把握好以上十二个注意点,能够有效避免实战中的各种&ldquo;坑&rdquo;,帮助技术在走出实验室,走向落地应用时,能发挥更加健壮、强大的作用</span></p>
  97. </div>
  98. <!--Faq Section-->
  99. </div>
  100. </div>
  101. </div>
  102. </div>
  103. <!--Sidebar Side-->
  104. <div class="sidebar-side sticky-container col-lg-4 col-md-12 col-sm-12">
  105. <aside class="sidebar">
  106. <div class="inner sticky-box">
  107. <!--Blog Category Widget-->
  108. <div class="sidebar-widget sidebar-blog-category">
  109. <ul class="blog-cat">
  110. <li class=""><a href="app.html">APP</a></li>
  111. <li class=""><a href="h5.html">H5</a></li>
  112. <li class=""><a href="mini-program.html">小程序</a></li>
  113. <li class=""><a href="web.html">WEB应用</a></li>
  114. <li class=""><a href="spider.html">网络爬虫</a></li>
  115. <li class="active "><a href="">AI、大数据</a></li>
  116. <li class=""><a href="crm.html">CRM</a></li>
  117. <li class=""><a href="erp.html">ERP</a></li>
  118. <li class=""><a href="oa.html">OA</a></li>
  119. <li class=""><a href="official-website.html">企业官网</a></li>
  120. <li class=""><a href="marketing-website.html">营销网站</a></li>
  121. <li class=""><a href="other-website.html">其他网站</a></li>
  122. </ul>
  123. </div>
  124. <!--Donate Help Widget-->
  125. <div class="sidebar-widget donate-help-widget">
  126. <div class="inner-box">
  127. <h2>如果需要我们的帮助</h2>
  128. <div class="text">
  129. 您可以通过电话或者在线聊天的方式找我们,我们将竭诚为您服务!
  130. </div>
  131. <a href="contact.html" class="theme-btn btn-style-three">联系我们</a>
  132. </div>
  133. </div>
  134. <!--Brochure-->
  135. <div class="sidebar-widget-two brochure-widget">
  136. <div class="brochure-box">
  137. <div class="inner">
  138. <span class="icon flaticon-phone-call"></span>
  139. <div class="text">173-4731-5117</div>
  140. </div>
  141. <a href="tel:17347315117" class="overlay-link"></a>
  142. </div>
  143. </div>
  144. </div>
  145. </aside>
  146. </div>
  147. <!--Content Side-->
  148. </div>
  149. </div>
  150. </div>
  151. <!--Subscribe Style One-->
  152. <section class="subscribe-style-one alternate">
  153. <div class="auto-container">
  154. <div class="row clearfix">
  155. <div class="col-lg-5 col-md-12 col-sm-12">
  156. <h2>现在开始沟通需求</h2>
  157. <div class="text">我们会对您的信息做加密处理,保证您的数据安全,我们不会将您的信息泄露给任何第三方!</div>
  158. </div>
  159. <div class="col-lg-7 col-md-12 col-sm-12">
  160. <form method="post" action="/contact" id="contact-form">
  161. <input type="hidden" name="_token" value="2pTP5xkN7Kg1MeiG8s8sTnd4qkhD5ke7FqrbqsXg"> <div class="form-group clearfix">
  162. <input type="tel" name="phone" value="" placeholder="输入手机号" required="">
  163. <button type="submit" class="theme-btn btn-style-one">提交</button>
  164. </div>
  165. </form>
  166. </div>
  167. </div>
  168. </div>
  169. </section>
  170. <!--End Subscribe Style One-->
  171. <!--Main Footer-->
  172. <div class="footer">
  173. <div class="footer-top">
  174. <div class="wrap">
  175. <div class="footer-block w-29">
  176. <div class="logo">
  177. <img src="static/image/login-1.png" alt="login">
  178. </div>
  179. <span class="partner" style="margin-left: -38px;">让APP开发变得更简单</span>
  180. </div>
  181. <div class="footer-block w-25 " style="width: 100%; margin: 0 auto;">
  182. <div class="footer-tit col-md-12 col-sm-12 col-xs-12" >服务热线</div>
  183. <div class="hot-line col-md-12 col-sm-12 col-xs-12">173-4731-5117</div>
  184. <p class="col-md-12 col-sm-12 col-xs-12"><span>售前咨询(09:00-24:00)</span></p>
  185. </div>
  186. <div class="footer-block w-34">
  187. <div class="footer-tit">联系我们</div>
  188. <p style="margin-bottom: 10px;">长沙市岳麓区枫林三路喜地大厦802室 </p>
  189. <p class="mb-10">
  190. <a href="tencent://message/?uin=2065525518&Site=&Menu=yes" target="_blank">
  191. 企业QQ:23933725
  192. <span class="foot-qq"></span>
  193. </a>
  194. </p>
  195. <p>企业邮箱:23933725@qq.com</p>
  196. <p>电话咨询:17347315117</p>
  197. </div>
  198. <div class="footer-block w-12">
  199. <div class="footer-tit">加官方微信</div>
  200. <div class="footer-ewm">
  201. <img src="static/image/erwei.jpg" alt="ewm">
  202. </div>
  203. <p class="ml">免费咨询报价及方案</p>
  204. </div>
  205. <div class="clear"></div>
  206. </div>
  207. </div>
  208. <div class="footer-bottom" style="height: auto;padding:10px 0;line-height: 20px;">
  209. <p>
  210. Copyright@2021-2022 湘ICP备2022000793号 湖南铭迅网络科技有限公司 版权所有 |
  211. </p>
  212. <ul class="ulbox">
  213. <li><a href="about.html">关于我们 |</a></li>
  214. <li><a href="faq.html">新闻资讯 |</a></li>
  215. <li><a href="contact.html">联系我们 |</a></li>
  216. <li><a href="map.html">网站地图 |</a></li>
  217. </ul>
  218. <ul class="ulboxs">
  219. <li>友情链接 :</li>
  220. <!-- <li><a>铺好租</a></li>
  221. <li><a>百万网址导航</a></li>
  222. <li><a>广州保洁公司</a></li>
  223. <li><a>九江公司</a></li>
  224. <li><a>长沙装修公司</a></li>
  225. <li><a>大集商铺网</a></li>
  226. <li><a>长沙门面出租</a></li> -->
  227. </ul>
  228. <div class="h40"></div>
  229. <!--End Footer Bottom-->
  230. <script>
  231. // var _hmt = _hmt || [];
  232. // (function() {
  233. // var hm = document.createElement("script");
  234. // hm.src = "https://hm.baidu.com/hm.js?7c8206c3cbaf6b515a6a72f50fbb4632";
  235. // var s = document.getElementsByTagName("script")[0];
  236. // s.parentNode.insertBefore(hm, s);
  237. // })();
  238. </script>
  239. </footer>
  240. </div>
  241. </div>
  242. <!--End pagewrapper-->
  243. <!--Scroll to top-->
  244. <div class="scroll-to-top scroll-to-target" data-target="html"><span class="fa fa-arrow-up"></span></div>
  245. <script src="static/js/jquery.js"></script>
  246. <script src="static/js/popper.min.js"></script>
  247. <script src="static/js/bootstrap.min.js"></script>
  248. <script src="static/js/jquery.mCustomScrollbar.concat.min.js"></script>
  249. <script src="static/js/jquery.fancybox.js"></script>
  250. <script src="static/js/appear.js"></script>
  251. <script src="static/js/owl.js"></script>
  252. <script src="static/js/sticky.js"></script>
  253. <script src="static/js/wow.js"></script>
  254. <script src="static/js/slick.js"></script>
  255. <script src="static/js/jquery.validate.min.js"></script>
  256. <script src="static/js/additional-methods.js"></script>
  257. <script src="static/js/messages_zh.min.js"></script>
  258. <script src="static/js/jquery-ui.js"></script>
  259. <script src="static/js/script.js"></script>
  260. </body>
  261. </html>