Skip to content

F
faiss
Switch branch/tag
Find file
BlameHistoryPermalink
  • Lucas Hosseini's avatar
    Facebook sync (May 2019) + relicense (#838) · a8118acb
    Lucas Hosseini authored 
    Changelog:

- changed license: BSD+Patents -> MIT
- propagates exceptions raised in sub-indexes of IndexShards and IndexReplicas
- support for searching several inverted lists in parallel (parallel_mode != 0)
- better support for PQ codes where nbit != 8 or 16
- IVFSpectralHash implementation: spectral hash codes inside an IVF
- 6-bit per component scalar quantizer (4 and 8 bit were already supported)
- combinations of inverted lists: HStackInvertedLists and VStackInvertedLists
- configurable number of threads for OnDiskInvertedLists prefetching (including 0=no prefetch)
- more test and demo code compatible with Python 3 (print with parentheses)
- refactored benchmark code: data loading is now in a single file
    Unverified
    a8118acb
PolysemousTraining_8h_source.html 40.2 KB
EditWeb IDE
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
<html xmlns="http://www.w3.org/1999/xhtml">
<head>
<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
<meta http-equiv="X-UA-Compatible" content="IE=9"/>
<meta name="generator" content="Doxygen 1.8.5"/>
<title>Faiss: /data/users/hoss/faiss/PolysemousTraining.h Source File</title>
<link href="tabs.css" rel="stylesheet" type="text/css"/>
<script type="text/javascript" src="jquery.js"></script>
<script type="text/javascript" src="dynsections.js"></script>
<link href="search/search.css" rel="stylesheet" type="text/css"/>
<script type="text/javascript" src="search/search.js"></script>
<script type="text/javascript">
  $(document).ready(function() { searchBox.OnSelectItem(0); });
</script>
<link href="doxygen.css" rel="stylesheet" type="text/css" />
</head>
<body>
<div id="top"><!-- do not remove this div, it is closed by doxygen! -->
<div id="titlearea">
<table cellspacing="0" cellpadding="0">
 <tbody>
 <tr style="height: 56px;">
  <td style="padding-left: 0.5em;">
   <div id="projectname">Faiss
   </div>
  </td>
 </tr>
 </tbody>
</table>
</div>
<!-- end header part -->
<!-- Generated by Doxygen 1.8.5 -->
<script type="text/javascript">
var searchBox = new SearchBox("searchBox", "search",false,'Search');
</script>
  <div id="navrow1" class="tabs">
    <ul class="tablist">
      <li><a href="index.html"><span>Main&#160;Page</span></a></li>
      <li><a href="namespaces.html"><span>Namespaces</span></a></li>
      <li><a href="annotated.html"><span>Classes</span></a></li>
      <li class="current"><a href="files.html"><span>Files</span></a></li>
      <li>
        <div id="MSearchBox" class="MSearchBoxInactive">
        <span class="left">
          <img id="MSearchSelect" src="search/mag_sel.png"
               onmouseover="return searchBox.OnSearchSelectShow()"
               onmouseout="return searchBox.OnSearchSelectHide()"
               alt=""/>
          <input type="text" id="MSearchField" value="Search" accesskey="S"
               onfocus="searchBox.OnSearchFieldFocus(true)" 
               onblur="searchBox.OnSearchFieldFocus(false)" 
               onkeyup="searchBox.OnSearchFieldChange(event)"/>
          </span><span class="right">
            <a id="MSearchClose" href="javascript:searchBox.CloseResultsWindow()"><img id="MSearchCloseImg" border="0" src="search/close.png" alt=""/></a>
          </span>
        </div>
      </li>
    </ul>
  </div>
  <div id="navrow2" class="tabs2">
    <ul class="tablist">
      <li><a href="files.html"><span>File&#160;List</span></a></li>
    </ul>
  </div>
</div><!-- top -->
<!-- window showing the filter options -->
<div id="MSearchSelectWindow"
     onmouseover="return searchBox.OnSearchSelectShow()"
     onmouseout="return searchBox.OnSearchSelectHide()"
     onkeydown="return searchBox.OnSearchSelectKey(event)">
<a class="SelectItem" href="javascript:void(0)" onclick="searchBox.OnSelectItem(0)"><span class="SelectionMark">&#160;</span>All</a><a class="SelectItem" href="javascript:void(0)" onclick="searchBox.OnSelectItem(1)"><span class="SelectionMark">&#160;</span>Classes</a><a class="SelectItem" href="javascript:void(0)" onclick="searchBox.OnSelectItem(2)"><span class="SelectionMark">&#160;</span>Namespaces</a><a class="SelectItem" href="javascript:void(0)" onclick="searchBox.OnSelectItem(3)"><span class="SelectionMark">&#160;</span>Functions</a><a class="SelectItem" href="javascript:void(0)" onclick="searchBox.OnSelectItem(4)"><span class="SelectionMark">&#160;</span>Variables</a><a class="SelectItem" href="javascript:void(0)" onclick="searchBox.OnSelectItem(5)"><span class="SelectionMark">&#160;</span>Typedefs</a><a class="SelectItem" href="javascript:void(0)" onclick="searchBox.OnSelectItem(6)"><span class="SelectionMark">&#160;</span>Enumerations</a><a class="SelectItem" href="javascript:void(0)" onclick="searchBox.OnSelectItem(7)"><span class="SelectionMark">&#160;</span>Enumerator</a><a class="SelectItem" href="javascript:void(0)" onclick="searchBox.OnSelectItem(8)"><span class="SelectionMark">&#160;</span>Friends</a></div>

<!-- iframe showing the search results (closed by default) -->
<div id="MSearchResultsWindow">
<iframe src="javascript:void(0)" frameborder="0" 
        name="MSearchResults" id="MSearchResults">
</iframe>
</div>

<div class="header">
  <div class="headertitle">
<div class="title">/data/users/hoss/faiss/PolysemousTraining.h</div>  </div>
</div><!--header-->
<div class="contents">
<div class="fragment"><div class="line"><a name="l00001"></a><span class="lineno">    1</span>&#160;<span class="comment">/**</span></div>
<div class="line"><a name="l00002"></a><span class="lineno">    2</span>&#160;<span class="comment"> * Copyright (c) Facebook, Inc. and its affiliates.</span></div>
<div class="line"><a name="l00003"></a><span class="lineno">    3</span>&#160;<span class="comment"> *</span></div>
<div class="line"><a name="l00004"></a><span class="lineno">    4</span>&#160;<span class="comment"> * This source code is licensed under the MIT license found in the</span></div>
<div class="line"><a name="l00005"></a><span class="lineno">    5</span>&#160;<span class="comment"> * LICENSE file in the root directory of this source tree.</span></div>
<div class="line"><a name="l00006"></a><span class="lineno">    6</span>&#160;<span class="comment"> */</span></div>
<div class="line"><a name="l00007"></a><span class="lineno">    7</span>&#160;</div>
<div class="line"><a name="l00008"></a><span class="lineno">    8</span>&#160;<span class="comment">// -*- c++ -*-</span></div>
<div class="line"><a name="l00009"></a><span class="lineno">    9</span>&#160;</div>
<div class="line"><a name="l00010"></a><span class="lineno">   10</span>&#160;<span class="preprocessor">#ifndef FAISS_POLYSEMOUS_TRAINING_INCLUDED</span></div>
<div class="line"><a name="l00011"></a><span class="lineno">   11</span>&#160;<span class="preprocessor"></span><span class="preprocessor">#define FAISS_POLYSEMOUS_TRAINING_INCLUDED</span></div>
<div class="line"><a name="l00012"></a><span class="lineno">   12</span>&#160;<span class="preprocessor"></span></div>
<div class="line"><a name="l00013"></a><span class="lineno">   13</span>&#160;</div>
<div class="line"><a name="l00014"></a><span class="lineno">   14</span>&#160;<span class="preprocessor">#include &quot;ProductQuantizer.h&quot;</span></div>
<div class="line"><a name="l00015"></a><span class="lineno">   15</span>&#160;</div>
<div class="line"><a name="l00016"></a><span class="lineno">   16</span>&#160;</div>
<div class="line"><a name="l00017"></a><span class="lineno">   17</span>&#160;<span class="keyword">namespace </span>faiss {</div>
<div class="line"><a name="l00018"></a><span class="lineno">   18</span>&#160;</div>
<div class="line"><a name="l00019"></a><span class="lineno">   19</span>&#160;<span class="comment"></span></div>
<div class="line"><a name="l00020"></a><span class="lineno">   20</span>&#160;<span class="comment">/// parameters used for the simulated annealing method</span></div>
<div class="line"><a name="l00021"></a><span class="lineno"><a class="line" href="structfaiss_1_1SimulatedAnnealingParameters.html">   21</a></span>&#160;<span class="comment"></span><span class="keyword">struct </span><a class="code" href="structfaiss_1_1SimulatedAnnealingParameters.html">SimulatedAnnealingParameters</a> {</div>
<div class="line"><a name="l00022"></a><span class="lineno">   22</span>&#160;</div>
<div class="line"><a name="l00023"></a><span class="lineno">   23</span>&#160;    <span class="comment">// optimization parameters</span></div>
<div class="line"><a name="l00024"></a><span class="lineno">   24</span>&#160;    <span class="keywordtype">double</span> init_temperature;   <span class="comment">// init probaility of accepting a bad swap</span></div>
<div class="line"><a name="l00025"></a><span class="lineno">   25</span>&#160;    <span class="keywordtype">double</span> temperature_decay;  <span class="comment">// at each iteration the temp is multiplied by this</span></div>
<div class="line"><a name="l00026"></a><span class="lineno">   26</span>&#160;    <span class="keywordtype">int</span> n_iter; <span class="comment">// nb of iterations</span></div>
<div class="line"><a name="l00027"></a><span class="lineno">   27</span>&#160;    <span class="keywordtype">int</span> n_redo; <span class="comment">// nb of runs of the simulation</span></div>
<div class="line"><a name="l00028"></a><span class="lineno">   28</span>&#160;    <span class="keywordtype">int</span> seed;   <span class="comment">// random seed</span></div>
<div class="line"><a name="l00029"></a><span class="lineno">   29</span>&#160;    <span class="keywordtype">int</span> verbose;</div>
<div class="line"><a name="l00030"></a><span class="lineno">   30</span>&#160;    <span class="keywordtype">bool</span> only_bit_flips; <span class="comment">// restrict permutation changes to bit flips</span></div>
<div class="line"><a name="l00031"></a><span class="lineno">   31</span>&#160;    <span class="keywordtype">bool</span> init_random; <span class="comment">// intialize with a random permutation (not identity)</span></div>
<div class="line"><a name="l00032"></a><span class="lineno">   32</span>&#160;</div>
<div class="line"><a name="l00033"></a><span class="lineno">   33</span>&#160;    <span class="comment">// set reasonable defaults</span></div>
<div class="line"><a name="l00034"></a><span class="lineno">   34</span>&#160;    <a class="code" href="structfaiss_1_1SimulatedAnnealingParameters.html">SimulatedAnnealingParameters</a> ();</div>
<div class="line"><a name="l00035"></a><span class="lineno">   35</span>&#160;</div>
<div class="line"><a name="l00036"></a><span class="lineno">   36</span>&#160;};</div>
<div class="line"><a name="l00037"></a><span class="lineno">   37</span>&#160;</div>
<div class="line"><a name="l00038"></a><span class="lineno">   38</span>&#160;<span class="comment"></span></div>
<div class="line"><a name="l00039"></a><span class="lineno">   39</span>&#160;<span class="comment">/// abstract class for the loss function</span></div>
<div class="line"><a name="l00040"></a><span class="lineno"><a class="line" href="structfaiss_1_1PermutationObjective.html">   40</a></span>&#160;<span class="comment"></span><span class="keyword">struct </span><a class="code" href="structfaiss_1_1PermutationObjective.html">PermutationObjective</a> {</div>
<div class="line"><a name="l00041"></a><span class="lineno">   41</span>&#160;</div>
<div class="line"><a name="l00042"></a><span class="lineno">   42</span>&#160;    <span class="keywordtype">int</span> n;</div>
<div class="line"><a name="l00043"></a><span class="lineno">   43</span>&#160;</div>
<div class="line"><a name="l00044"></a><span class="lineno">   44</span>&#160;    <span class="keyword">virtual</span> <span class="keywordtype">double</span> compute_cost (<span class="keyword">const</span> <span class="keywordtype">int</span> *perm) <span class="keyword">const</span> = 0;</div>
<div class="line"><a name="l00045"></a><span class="lineno">   45</span>&#160;</div>
<div class="line"><a name="l00046"></a><span class="lineno">   46</span>&#160;    <span class="comment">// what would the cost update be if iw and jw were swapped?</span></div>
<div class="line"><a name="l00047"></a><span class="lineno">   47</span>&#160;    <span class="comment">// default implementation just computes both and computes the difference</span></div>
<div class="line"><a name="l00048"></a><span class="lineno">   48</span>&#160;    <span class="keyword">virtual</span> <span class="keywordtype">double</span> cost_update (<span class="keyword">const</span> <span class="keywordtype">int</span> *perm, <span class="keywordtype">int</span> iw, <span class="keywordtype">int</span> jw) <span class="keyword">const</span>;</div>
<div class="line"><a name="l00049"></a><span class="lineno">   49</span>&#160;</div>
<div class="line"><a name="l00050"></a><span class="lineno">   50</span>&#160;    <span class="keyword">virtual</span> ~<a class="code" href="structfaiss_1_1PermutationObjective.html">PermutationObjective</a> () {}</div>
<div class="line"><a name="l00051"></a><span class="lineno">   51</span>&#160;};</div>
<div class="line"><a name="l00052"></a><span class="lineno">   52</span>&#160;</div>
<div class="line"><a name="l00053"></a><span class="lineno">   53</span>&#160;</div>
<div class="line"><a name="l00054"></a><span class="lineno"><a class="line" href="structfaiss_1_1ReproduceDistancesObjective.html">   54</a></span>&#160;<span class="keyword">struct </span><a class="code" href="structfaiss_1_1ReproduceDistancesObjective.html">ReproduceDistancesObjective</a> : <a class="code" href="structfaiss_1_1PermutationObjective.html">PermutationObjective</a> {</div>
<div class="line"><a name="l00055"></a><span class="lineno">   55</span>&#160;</div>
<div class="line"><a name="l00056"></a><span class="lineno">   56</span>&#160;    <span class="keywordtype">double</span> dis_weight_factor;</div>
<div class="line"><a name="l00057"></a><span class="lineno">   57</span>&#160;</div>
<div class="line"><a name="l00058"></a><span class="lineno">   58</span>&#160;    <span class="keyword">static</span> <span class="keywordtype">double</span> sqr (<span class="keywordtype">double</span> x) { <span class="keywordflow">return</span> x * x; }</div>
<div class="line"><a name="l00059"></a><span class="lineno">   59</span>&#160;</div>
<div class="line"><a name="l00060"></a><span class="lineno">   60</span>&#160;    <span class="comment">// weihgting of distances: it is more important to reproduce small</span></div>
<div class="line"><a name="l00061"></a><span class="lineno">   61</span>&#160;    <span class="comment">// distances well</span></div>
<div class="line"><a name="l00062"></a><span class="lineno">   62</span>&#160;    <span class="keywordtype">double</span> dis_weight (<span class="keywordtype">double</span> x) <span class="keyword">const</span>;</div>
<div class="line"><a name="l00063"></a><span class="lineno">   63</span>&#160;</div>
<div class="line"><a name="l00064"></a><span class="lineno"><a class="line" href="structfaiss_1_1ReproduceDistancesObjective.html#a6681846eb7751b32a907d075fff6770b">   64</a></span>&#160;    std::vector&lt;double&gt; <a class="code" href="structfaiss_1_1ReproduceDistancesObjective.html#a6681846eb7751b32a907d075fff6770b">source_dis</a>; <span class="comment">///&lt; &quot;real&quot; corrected distances (size n^2)</span></div>
<div class="line"><a name="l00065"></a><span class="lineno"><a class="line" href="structfaiss_1_1ReproduceDistancesObjective.html#ac99d739f81e13e841ba7d81d90629aef">   65</a></span>&#160;<span class="comment"></span>    <span class="keyword">const</span> <span class="keywordtype">double</span> *      <a class="code" href="structfaiss_1_1ReproduceDistancesObjective.html#ac99d739f81e13e841ba7d81d90629aef">target_dis</a>; <span class="comment">///&lt; wanted distances (size n^2)</span></div>
<div class="line"><a name="l00066"></a><span class="lineno"><a class="line" href="structfaiss_1_1ReproduceDistancesObjective.html#a982332b44ff32b9ae9694e374acd0aff">   66</a></span>&#160;<span class="comment"></span>    std::vector&lt;double&gt; <a class="code" href="structfaiss_1_1ReproduceDistancesObjective.html#a982332b44ff32b9ae9694e374acd0aff">weights</a>;    <span class="comment">///&lt; weights for each distance (size n^2)</span></div>
<div class="line"><a name="l00067"></a><span class="lineno">   67</span>&#160;<span class="comment"></span></div>
<div class="line"><a name="l00068"></a><span class="lineno">   68</span>&#160;    <span class="keywordtype">double</span> get_source_dis (<span class="keywordtype">int</span> i, <span class="keywordtype">int</span> j) <span class="keyword">const</span>;</div>
<div class="line"><a name="l00069"></a><span class="lineno">   69</span>&#160;</div>
<div class="line"><a name="l00070"></a><span class="lineno">   70</span>&#160;    <span class="comment">// cost = quadratic difference between actual distance and Hamming distance</span></div>
<div class="line"><a name="l00071"></a><span class="lineno">   71</span>&#160;    <span class="keywordtype">double</span> compute_cost(<span class="keyword">const</span> <span class="keywordtype">int</span>* perm) <span class="keyword">const override</span>;</div>
<div class="line"><a name="l00072"></a><span class="lineno">   72</span>&#160;</div>
<div class="line"><a name="l00073"></a><span class="lineno">   73</span>&#160;    <span class="comment">// what would the cost update be if iw and jw were swapped?</span></div>
<div class="line"><a name="l00074"></a><span class="lineno">   74</span>&#160;    <span class="comment">// computed in O(n) instead of O(n^2) for the full re-computation</span></div>
<div class="line"><a name="l00075"></a><span class="lineno">   75</span>&#160;    <span class="keywordtype">double</span> cost_update(<span class="keyword">const</span> <span class="keywordtype">int</span>* perm, <span class="keywordtype">int</span> iw, <span class="keywordtype">int</span> jw) <span class="keyword">const override</span>;</div>
<div class="line"><a name="l00076"></a><span class="lineno">   76</span>&#160;</div>
<div class="line"><a name="l00077"></a><span class="lineno">   77</span>&#160;    <a class="code" href="structfaiss_1_1ReproduceDistancesObjective.html">ReproduceDistancesObjective</a> (</div>
<div class="line"><a name="l00078"></a><span class="lineno">   78</span>&#160;           <span class="keywordtype">int</span> n,</div>
<div class="line"><a name="l00079"></a><span class="lineno">   79</span>&#160;           <span class="keyword">const</span> <span class="keywordtype">double</span> *source_dis_in,</div>
<div class="line"><a name="l00080"></a><span class="lineno">   80</span>&#160;           <span class="keyword">const</span> <span class="keywordtype">double</span> *target_dis_in,</div>
<div class="line"><a name="l00081"></a><span class="lineno">   81</span>&#160;           <span class="keywordtype">double</span> dis_weight_factor);</div>
<div class="line"><a name="l00082"></a><span class="lineno">   82</span>&#160;</div>
<div class="line"><a name="l00083"></a><span class="lineno">   83</span>&#160;    <span class="keyword">static</span> <span class="keywordtype">void</span> compute_mean_stdev (<span class="keyword">const</span> <span class="keywordtype">double</span> *tab, <span class="keywordtype">size_t</span> n2,</div>
<div class="line"><a name="l00084"></a><span class="lineno">   84</span>&#160;                                    <span class="keywordtype">double</span> *mean_out, <span class="keywordtype">double</span> *stddev_out);</div>
<div class="line"><a name="l00085"></a><span class="lineno">   85</span>&#160;</div>
<div class="line"><a name="l00086"></a><span class="lineno">   86</span>&#160;    <span class="keywordtype">void</span> set_affine_target_dis (<span class="keyword">const</span> <span class="keywordtype">double</span> *source_dis_in);</div>
<div class="line"><a name="l00087"></a><span class="lineno">   87</span>&#160;</div>
<div class="line"><a name="l00088"></a><span class="lineno">   88</span>&#160;    ~<a class="code" href="structfaiss_1_1ReproduceDistancesObjective.html">ReproduceDistancesObjective</a>()<span class="keyword"> override </span>{}</div>
<div class="line"><a name="l00089"></a><span class="lineno">   89</span>&#160;};</div>
<div class="line"><a name="l00090"></a><span class="lineno">   90</span>&#160;</div>
<div class="line"><a name="l00091"></a><span class="lineno">   91</span>&#160;<span class="keyword">struct </span>RandomGenerator;</div>
<div class="line"><a name="l00092"></a><span class="lineno">   92</span>&#160;<span class="comment"></span></div>
<div class="line"><a name="l00093"></a><span class="lineno">   93</span>&#160;<span class="comment">/// Simulated annealing optimization algorithm for permutations.</span></div>
<div class="line"><a name="l00094"></a><span class="lineno"><a class="line" href="structfaiss_1_1SimulatedAnnealingOptimizer.html">   94</a></span>&#160;<span class="comment"></span> <span class="keyword">struct </span><a class="code" href="structfaiss_1_1SimulatedAnnealingOptimizer.html">SimulatedAnnealingOptimizer</a>: <a class="code" href="structfaiss_1_1SimulatedAnnealingParameters.html">SimulatedAnnealingParameters</a> {</div>
<div class="line"><a name="l00095"></a><span class="lineno">   95</span>&#160;</div>
<div class="line"><a name="l00096"></a><span class="lineno">   96</span>&#160;    <a class="code" href="structfaiss_1_1PermutationObjective.html">PermutationObjective</a> *obj;</div>
<div class="line"><a name="l00097"></a><span class="lineno"><a class="line" href="structfaiss_1_1SimulatedAnnealingOptimizer.html#a90edfc401372aa1a2a832b403e26bfd1">   97</a></span>&#160;    <span class="keywordtype">int</span> <a class="code" href="structfaiss_1_1SimulatedAnnealingOptimizer.html#a90edfc401372aa1a2a832b403e26bfd1">n</a>;         <span class="comment">///&lt; size of the permutation</span></div>
<div class="line"><a name="l00098"></a><span class="lineno">   98</span>&#160;<span class="comment"></span>    FILE *logfile; <span class="comment">/// logs values of the cost function</span></div>
<div class="line"><a name="l00099"></a><span class="lineno">   99</span>&#160;<span class="comment"></span></div>
<div class="line"><a name="l00100"></a><span class="lineno">  100</span>&#160;    <a class="code" href="structfaiss_1_1SimulatedAnnealingOptimizer.html#ab258eaf892742e37dd2c3bfd3b1c39e1">SimulatedAnnealingOptimizer</a> (<a class="code" href="structfaiss_1_1PermutationObjective.html">PermutationObjective</a> *obj,</div>
<div class="line"><a name="l00101"></a><span class="lineno">  101</span>&#160;                                 <span class="keyword">const</span> <a class="code" href="structfaiss_1_1SimulatedAnnealingParameters.html">SimulatedAnnealingParameters</a> &amp;p);</div>
<div class="line"><a name="l00102"></a><span class="lineno">  102</span>&#160;    <a class="code" href="structfaiss_1_1RandomGenerator.html">RandomGenerator</a> *rnd;</div>
<div class="line"><a name="l00103"></a><span class="lineno">  103</span>&#160;<span class="comment"></span></div>
<div class="line"><a name="l00104"></a><span class="lineno">  104</span>&#160;<span class="comment">    /// remember intial cost of optimization</span></div>
<div class="line"><a name="l00105"></a><span class="lineno"><a class="line" href="structfaiss_1_1SimulatedAnnealingOptimizer.html#aa8ff19f97482c6213b2a8ff3b190412b">  105</a></span>&#160;<span class="comment"></span>    <span class="keywordtype">double</span> <a class="code" href="structfaiss_1_1SimulatedAnnealingOptimizer.html#aa8ff19f97482c6213b2a8ff3b190412b">init_cost</a>;</div>
<div class="line"><a name="l00106"></a><span class="lineno">  106</span>&#160;</div>
<div class="line"><a name="l00107"></a><span class="lineno">  107</span>&#160;    <span class="comment">// main entry point. Perform the optimization loop, starting from</span></div>
<div class="line"><a name="l00108"></a><span class="lineno">  108</span>&#160;    <span class="comment">// and modifying permutation in-place</span></div>
<div class="line"><a name="l00109"></a><span class="lineno">  109</span>&#160;    <span class="keywordtype">double</span> optimize (<span class="keywordtype">int</span> *perm);</div>
<div class="line"><a name="l00110"></a><span class="lineno">  110</span>&#160;</div>
<div class="line"><a name="l00111"></a><span class="lineno">  111</span>&#160;    <span class="comment">// run the optimization and return the best result in best_perm</span></div>
<div class="line"><a name="l00112"></a><span class="lineno">  112</span>&#160;    <span class="keywordtype">double</span> run_optimization (<span class="keywordtype">int</span> * best_perm);</div>
<div class="line"><a name="l00113"></a><span class="lineno">  113</span>&#160;</div>
<div class="line"><a name="l00114"></a><span class="lineno">  114</span>&#160;    <span class="keyword">virtual</span> ~<a class="code" href="structfaiss_1_1SimulatedAnnealingOptimizer.html">SimulatedAnnealingOptimizer</a> ();</div>
<div class="line"><a name="l00115"></a><span class="lineno">  115</span>&#160;};</div>
<div class="line"><a name="l00116"></a><span class="lineno">  116</span>&#160;</div>
<div class="line"><a name="l00117"></a><span class="lineno">  117</span>&#160;</div>
<div class="line"><a name="l00118"></a><span class="lineno">  118</span>&#160;</div>
<div class="line"><a name="l00119"></a><span class="lineno">  119</span>&#160;<span class="comment"></span></div>
<div class="line"><a name="l00120"></a><span class="lineno">  120</span>&#160;<span class="comment">/// optimizes the order of indices in a ProductQuantizer</span></div>
<div class="line"><a name="l00121"></a><span class="lineno"><a class="line" href="structfaiss_1_1PolysemousTraining.html">  121</a></span>&#160;<span class="comment"></span><span class="keyword">struct </span><a class="code" href="structfaiss_1_1PolysemousTraining.html">PolysemousTraining</a>: <a class="code" href="structfaiss_1_1SimulatedAnnealingParameters.html">SimulatedAnnealingParameters</a> {</div>
<div class="line"><a name="l00122"></a><span class="lineno">  122</span>&#160;</div>
<div class="line"><a name="l00123"></a><span class="lineno"><a class="line" href="structfaiss_1_1PolysemousTraining.html#aeb5a523056697934972f95fd428f61ef">  123</a></span>&#160;    <span class="keyword">enum</span> <a class="code" href="structfaiss_1_1PolysemousTraining.html#aeb5a523056697934972f95fd428f61ef">Optimization_type_t</a> {</div>
<div class="line"><a name="l00124"></a><span class="lineno">  124</span>&#160;        OT_None,</div>
<div class="line"><a name="l00125"></a><span class="lineno"><a class="line" href="structfaiss_1_1PolysemousTraining.html#aeb5a523056697934972f95fd428f61efa4f1f59a0027c1662a7aaf66f2a78fb23">  125</a></span>&#160;        <a class="code" href="structfaiss_1_1PolysemousTraining.html#aeb5a523056697934972f95fd428f61efa4f1f59a0027c1662a7aaf66f2a78fb23">OT_ReproduceDistances_affine</a>,  <span class="comment">///&lt; default</span></div>
<div class="line"><a name="l00126"></a><span class="lineno"><a class="line" href="structfaiss_1_1PolysemousTraining.html#aeb5a523056697934972f95fd428f61efa514d33b85a240091cdd7b8efaf66a0b5">  126</a></span>&#160;<span class="comment"></span>        <a class="code" href="structfaiss_1_1PolysemousTraining.html#aeb5a523056697934972f95fd428f61efa514d33b85a240091cdd7b8efaf66a0b5">OT_Ranking_weighted_diff</a>  <span class="comment">/// same as _2, but use rank of y+ - rank of y-</span></div>
<div class="line"><a name="l00127"></a><span class="lineno">  127</span>&#160;<span class="comment"></span>    };</div>
<div class="line"><a name="l00128"></a><span class="lineno">  128</span>&#160;    <a class="code" href="structfaiss_1_1PolysemousTraining.html#aeb5a523056697934972f95fd428f61ef">Optimization_type_t</a> optimization_type;</div>
<div class="line"><a name="l00129"></a><span class="lineno">  129</span>&#160;</div>
<div class="line"><a name="l00130"></a><span class="lineno">  130</span>&#160;    <span class="comment">// use 1/4 of the training points for the optimization, with</span></div>
<div class="line"><a name="l00131"></a><span class="lineno">  131</span>&#160;    <span class="comment">// max. ntrain_permutation. If ntrain_permutation == 0: train on</span></div>
<div class="line"><a name="l00132"></a><span class="lineno">  132</span>&#160;    <span class="comment">// centroids</span></div>
<div class="line"><a name="l00133"></a><span class="lineno">  133</span>&#160;    <span class="keywordtype">int</span> ntrain_permutation;</div>
<div class="line"><a name="l00134"></a><span class="lineno">  134</span>&#160;    <span class="keywordtype">double</span> dis_weight_factor; <span class="comment">// decay of exp that weights distance loss</span></div>
<div class="line"><a name="l00135"></a><span class="lineno">  135</span>&#160;</div>
<div class="line"><a name="l00136"></a><span class="lineno">  136</span>&#160;    <span class="comment">// filename pattern for the logging of iterations</span></div>
<div class="line"><a name="l00137"></a><span class="lineno">  137</span>&#160;    std::string log_pattern;</div>
<div class="line"><a name="l00138"></a><span class="lineno">  138</span>&#160;</div>
<div class="line"><a name="l00139"></a><span class="lineno">  139</span>&#160;    <span class="comment">// sets default values</span></div>
<div class="line"><a name="l00140"></a><span class="lineno">  140</span>&#160;    <a class="code" href="structfaiss_1_1PolysemousTraining.html">PolysemousTraining</a> ();</div>
<div class="line"><a name="l00141"></a><span class="lineno">  141</span>&#160;<span class="comment"></span></div>
<div class="line"><a name="l00142"></a><span class="lineno">  142</span>&#160;<span class="comment">    /// reorder the centroids so that the Hamming distace becomes a</span></div>
<div class="line"><a name="l00143"></a><span class="lineno">  143</span>&#160;<span class="comment">    /// good approximation of the SDC distance (called by train)</span></div>
<div class="line"><a name="l00144"></a><span class="lineno">  144</span>&#160;<span class="comment"></span>    <span class="keywordtype">void</span> <a class="code" href="structfaiss_1_1PolysemousTraining.html#a995374ba2f0bc2738edae72ca9142f82">optimize_pq_for_hamming</a> (<a class="code" href="structfaiss_1_1ProductQuantizer.html">ProductQuantizer</a> &amp; pq,</div>
<div class="line"><a name="l00145"></a><span class="lineno">  145</span>&#160;                                  <span class="keywordtype">size_t</span> n, <span class="keyword">const</span> <span class="keywordtype">float</span> *x) <span class="keyword">const</span>;</div>
<div class="line"><a name="l00146"></a><span class="lineno">  146</span>&#160;<span class="comment"></span></div>
<div class="line"><a name="l00147"></a><span class="lineno">  147</span>&#160;<span class="comment">    /// called by optimize_pq_for_hamming</span></div>
<div class="line"><a name="l00148"></a><span class="lineno">  148</span>&#160;<span class="comment"></span>    <span class="keywordtype">void</span> <a class="code" href="structfaiss_1_1PolysemousTraining.html#a0ac61389b6fcc7f98e998d9d0cc2664a">optimize_ranking</a> (<a class="code" href="structfaiss_1_1ProductQuantizer.html">ProductQuantizer</a> &amp;pq, <span class="keywordtype">size_t</span> n, <span class="keyword">const</span> <span class="keywordtype">float</span> *x) <span class="keyword">const</span>;<span class="comment"></span></div>
<div class="line"><a name="l00149"></a><span class="lineno">  149</span>&#160;<span class="comment">    /// called by optimize_pq_for_hamming</span></div>
<div class="line"><a name="l00150"></a><span class="lineno">  150</span>&#160;<span class="comment"></span>    <span class="keywordtype">void</span> <a class="code" href="structfaiss_1_1PolysemousTraining.html#a6989e2b573d4368e7f863edbbb2f00b8">optimize_reproduce_distances</a> (<a class="code" href="structfaiss_1_1ProductQuantizer.html">ProductQuantizer</a> &amp;pq) <span class="keyword">const</span>;</div>
<div class="line"><a name="l00151"></a><span class="lineno">  151</span>&#160;</div>
<div class="line"><a name="l00152"></a><span class="lineno">  152</span>&#160;};</div>
<div class="line"><a name="l00153"></a><span class="lineno">  153</span>&#160;</div>
<div class="line"><a name="l00154"></a><span class="lineno">  154</span>&#160;</div>
<div class="line"><a name="l00155"></a><span class="lineno">  155</span>&#160;} <span class="comment">// namespace faiss</span></div>
<div class="line"><a name="l00156"></a><span class="lineno">  156</span>&#160;</div>
<div class="line"><a name="l00157"></a><span class="lineno">  157</span>&#160;</div>
<div class="line"><a name="l00158"></a><span class="lineno">  158</span>&#160;<span class="preprocessor">#endif</span></div>
<div class="ttc" id="structfaiss_1_1RandomGenerator_html"><div class="ttname"><a href="structfaiss_1_1RandomGenerator.html">faiss::RandomGenerator</a></div><div class="ttdoc">random generator that can be used in multithreaded contexts </div><div class="ttdef"><b>Definition:</b> <a href="utils_8h_source.html#l00047">utils.h:47</a></div></div>
<div class="ttc" id="structfaiss_1_1PolysemousTraining_html_aeb5a523056697934972f95fd428f61efa514d33b85a240091cdd7b8efaf66a0b5"><div class="ttname"><a href="structfaiss_1_1PolysemousTraining.html#aeb5a523056697934972f95fd428f61efa514d33b85a240091cdd7b8efaf66a0b5">faiss::PolysemousTraining::OT_Ranking_weighted_diff</a></div><div class="ttdoc">same as _2, but use rank of y+ - rank of y- </div><div class="ttdef"><b>Definition:</b> <a href="PolysemousTraining_8h_source.html#l00126">PolysemousTraining.h:126</a></div></div>
<div class="ttc" id="structfaiss_1_1SimulatedAnnealingOptimizer_html_ab258eaf892742e37dd2c3bfd3b1c39e1"><div class="ttname"><a href="structfaiss_1_1SimulatedAnnealingOptimizer.html#ab258eaf892742e37dd2c3bfd3b1c39e1">faiss::SimulatedAnnealingOptimizer::SimulatedAnnealingOptimizer</a></div><div class="ttdeci">SimulatedAnnealingOptimizer(PermutationObjective *obj, const SimulatedAnnealingParameters &amp;p)</div><div class="ttdoc">logs values of the cost function </div><div class="ttdef"><b>Definition:</b> <a href="PolysemousTraining_8cpp_source.html#l00068">PolysemousTraining.cpp:68</a></div></div>
<div class="ttc" id="structfaiss_1_1ReproduceDistancesObjective_html"><div class="ttname"><a href="structfaiss_1_1ReproduceDistancesObjective.html">faiss::ReproduceDistancesObjective</a></div><div class="ttdef"><b>Definition:</b> <a href="PolysemousTraining_8h_source.html#l00054">PolysemousTraining.h:54</a></div></div>
<div class="ttc" id="structfaiss_1_1SimulatedAnnealingOptimizer_html_a90edfc401372aa1a2a832b403e26bfd1"><div class="ttname"><a href="structfaiss_1_1SimulatedAnnealingOptimizer.html#a90edfc401372aa1a2a832b403e26bfd1">faiss::SimulatedAnnealingOptimizer::n</a></div><div class="ttdeci">int n</div><div class="ttdoc">size of the permutation </div><div class="ttdef"><b>Definition:</b> <a href="PolysemousTraining_8h_source.html#l00097">PolysemousTraining.h:97</a></div></div>
<div class="ttc" id="structfaiss_1_1ReproduceDistancesObjective_html_ac99d739f81e13e841ba7d81d90629aef"><div class="ttname"><a href="structfaiss_1_1ReproduceDistancesObjective.html#ac99d739f81e13e841ba7d81d90629aef">faiss::ReproduceDistancesObjective::target_dis</a></div><div class="ttdeci">const double * target_dis</div><div class="ttdoc">wanted distances (size n^2) </div><div class="ttdef"><b>Definition:</b> <a href="PolysemousTraining_8h_source.html#l00065">PolysemousTraining.h:65</a></div></div>
<div class="ttc" id="structfaiss_1_1SimulatedAnnealingOptimizer_html_aa8ff19f97482c6213b2a8ff3b190412b"><div class="ttname"><a href="structfaiss_1_1SimulatedAnnealingOptimizer.html#aa8ff19f97482c6213b2a8ff3b190412b">faiss::SimulatedAnnealingOptimizer::init_cost</a></div><div class="ttdeci">double init_cost</div><div class="ttdoc">remember intial cost of optimization </div><div class="ttdef"><b>Definition:</b> <a href="PolysemousTraining_8h_source.html#l00105">PolysemousTraining.h:105</a></div></div>
<div class="ttc" id="structfaiss_1_1PolysemousTraining_html_a0ac61389b6fcc7f98e998d9d0cc2664a"><div class="ttname"><a href="structfaiss_1_1PolysemousTraining.html#a0ac61389b6fcc7f98e998d9d0cc2664a">faiss::PolysemousTraining::optimize_ranking</a></div><div class="ttdeci">void optimize_ranking(ProductQuantizer &amp;pq, size_t n, const float *x) const </div><div class="ttdoc">called by optimize_pq_for_hamming </div><div class="ttdef"><b>Definition:</b> <a href="PolysemousTraining_8cpp_source.html#l00827">PolysemousTraining.cpp:827</a></div></div>
<div class="ttc" id="structfaiss_1_1PolysemousTraining_html"><div class="ttname"><a href="structfaiss_1_1PolysemousTraining.html">faiss::PolysemousTraining</a></div><div class="ttdoc">optimizes the order of indices in a ProductQuantizer </div><div class="ttdef"><b>Definition:</b> <a href="PolysemousTraining_8h_source.html#l00121">PolysemousTraining.h:121</a></div></div>
<div class="ttc" id="structfaiss_1_1ReproduceDistancesObjective_html_a982332b44ff32b9ae9694e374acd0aff"><div class="ttname"><a href="structfaiss_1_1ReproduceDistancesObjective.html#a982332b44ff32b9ae9694e374acd0aff">faiss::ReproduceDistancesObjective::weights</a></div><div class="ttdeci">std::vector&lt; double &gt; weights</div><div class="ttdoc">weights for each distance (size n^2) </div><div class="ttdef"><b>Definition:</b> <a href="PolysemousTraining_8h_source.html#l00066">PolysemousTraining.h:66</a></div></div>
<div class="ttc" id="structfaiss_1_1SimulatedAnnealingParameters_html"><div class="ttname"><a href="structfaiss_1_1SimulatedAnnealingParameters.html">faiss::SimulatedAnnealingParameters</a></div><div class="ttdoc">parameters used for the simulated annealing method </div><div class="ttdef"><b>Definition:</b> <a href="PolysemousTraining_8h_source.html#l00021">PolysemousTraining.h:21</a></div></div>
<div class="ttc" id="structfaiss_1_1PolysemousTraining_html_aeb5a523056697934972f95fd428f61efa4f1f59a0027c1662a7aaf66f2a78fb23"><div class="ttname"><a href="structfaiss_1_1PolysemousTraining.html#aeb5a523056697934972f95fd428f61efa4f1f59a0027c1662a7aaf66f2a78fb23">faiss::PolysemousTraining::OT_ReproduceDistances_affine</a></div><div class="ttdoc">default </div><div class="ttdef"><b>Definition:</b> <a href="PolysemousTraining_8h_source.html#l00125">PolysemousTraining.h:125</a></div></div>
<div class="ttc" id="structfaiss_1_1PermutationObjective_html"><div class="ttname"><a href="structfaiss_1_1PermutationObjective.html">faiss::PermutationObjective</a></div><div class="ttdoc">abstract class for the loss function </div><div class="ttdef"><b>Definition:</b> <a href="PolysemousTraining_8h_source.html#l00040">PolysemousTraining.h:40</a></div></div>
<div class="ttc" id="structfaiss_1_1PolysemousTraining_html_aeb5a523056697934972f95fd428f61ef"><div class="ttname"><a href="structfaiss_1_1PolysemousTraining.html#aeb5a523056697934972f95fd428f61ef">faiss::PolysemousTraining::Optimization_type_t</a></div><div class="ttdeci">Optimization_type_t</div><div class="ttdef"><b>Definition:</b> <a href="PolysemousTraining_8h_source.html#l00123">PolysemousTraining.h:123</a></div></div>
<div class="ttc" id="structfaiss_1_1ProductQuantizer_html"><div class="ttname"><a href="structfaiss_1_1ProductQuantizer.html">faiss::ProductQuantizer</a></div><div class="ttdef"><b>Definition:</b> <a href="ProductQuantizer_8h_source.html#l00023">ProductQuantizer.h:23</a></div></div>
<div class="ttc" id="structfaiss_1_1ReproduceDistancesObjective_html_a6681846eb7751b32a907d075fff6770b"><div class="ttname"><a href="structfaiss_1_1ReproduceDistancesObjective.html#a6681846eb7751b32a907d075fff6770b">faiss::ReproduceDistancesObjective::source_dis</a></div><div class="ttdeci">std::vector&lt; double &gt; source_dis</div><div class="ttdoc">&amp;quot;real&amp;quot; corrected distances (size n^2) </div><div class="ttdef"><b>Definition:</b> <a href="PolysemousTraining_8h_source.html#l00064">PolysemousTraining.h:64</a></div></div>
<div class="ttc" id="structfaiss_1_1PolysemousTraining_html_a6989e2b573d4368e7f863edbbb2f00b8"><div class="ttname"><a href="structfaiss_1_1PolysemousTraining.html#a6989e2b573d4368e7f863edbbb2f00b8">faiss::PolysemousTraining::optimize_reproduce_distances</a></div><div class="ttdeci">void optimize_reproduce_distances(ProductQuantizer &amp;pq) const </div><div class="ttdoc">called by optimize_pq_for_hamming </div><div class="ttdef"><b>Definition:</b> <a href="PolysemousTraining_8cpp_source.html#l00764">PolysemousTraining.cpp:764</a></div></div>
<div class="ttc" id="structfaiss_1_1PolysemousTraining_html_a995374ba2f0bc2738edae72ca9142f82"><div class="ttname"><a href="structfaiss_1_1PolysemousTraining.html#a995374ba2f0bc2738edae72ca9142f82">faiss::PolysemousTraining::optimize_pq_for_hamming</a></div><div class="ttdeci">void optimize_pq_for_hamming(ProductQuantizer &amp;pq, size_t n, const float *x) const </div><div class="ttdef"><b>Definition:</b> <a href="PolysemousTraining_8cpp_source.html#l00934">PolysemousTraining.cpp:934</a></div></div>
<div class="ttc" id="structfaiss_1_1SimulatedAnnealingOptimizer_html"><div class="ttname"><a href="structfaiss_1_1SimulatedAnnealingOptimizer.html">faiss::SimulatedAnnealingOptimizer</a></div><div class="ttdoc">Simulated annealing optimization algorithm for permutations. </div><div class="ttdef"><b>Definition:</b> <a href="PolysemousTraining_8h_source.html#l00094">PolysemousTraining.h:94</a></div></div>
</div><!-- fragment --></div><!-- contents -->
<!-- start footer part -->
<hr class="footer"/><address class="footer"><small>
Generated by &#160;<a href="http://www.doxygen.org/index.html">
<img class="footer" src="doxygen.png" alt="doxygen"/>
</a> 1.8.5
</small></address>
</body>
</html>