-
Notifications
You must be signed in to change notification settings - Fork 0
/
index.html
156 lines (138 loc) · 9.62 KB
/
index.html
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
<!doctype html>
<html lang="en">
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1, minimum-scale=1" />
<meta name="generator" content="pdoc 0.7.5" />
<title>dpm API documentation</title>
<meta name="description" content="Weighted Dirichlet Process Gaussian Mixture Model
This repo extends …" />
<link href='https://cdnjs.cloudflare.com/ajax/libs/normalize/8.0.0/normalize.min.css' rel='stylesheet'>
<link href='https://cdnjs.cloudflare.com/ajax/libs/10up-sanitize.css/8.0.0/sanitize.min.css' rel='stylesheet'>
<link href="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/9.12.0/styles/github.min.css" rel="stylesheet">
<style>.flex{display:flex !important}body{line-height:1.5em}#content{padding:20px}#sidebar{padding:30px;overflow:hidden}.http-server-breadcrumbs{font-size:130%;margin:0 0 15px 0}#footer{font-size:.75em;padding:5px 30px;border-top:1px solid #ddd;text-align:right}#footer p{margin:0 0 0 1em;display:inline-block}#footer p:last-child{margin-right:30px}h1,h2,h3,h4,h5{font-weight:300}h1{font-size:2.5em;line-height:1.1em}h2{font-size:1.75em;margin:1em 0 .50em 0}h3{font-size:1.4em;margin:25px 0 10px 0}h4{margin:0;font-size:105%}a{color:#058;text-decoration:none;transition:color .3s ease-in-out}a:hover{color:#e82}.title code{font-weight:bold}h2[id^="header-"]{margin-top:2em}.ident{color:#900}pre code{background:#f8f8f8;font-size:.8em;line-height:1.4em}code{background:#f2f2f1;padding:1px 4px;overflow-wrap:break-word}h1 code{background:transparent}pre{background:#f8f8f8;border:0;border-top:1px solid #ccc;border-bottom:1px solid #ccc;margin:1em 0;padding:1ex}#http-server-module-list{display:flex;flex-flow:column}#http-server-module-list div{display:flex}#http-server-module-list dt{min-width:10%}#http-server-module-list p{margin-top:0}.toc ul,#index{list-style-type:none;margin:0;padding:0}#index code{background:transparent}#index h3{border-bottom:1px solid #ddd}#index ul{padding:0}#index h4{font-weight:bold}#index h4 + ul{margin-bottom:.6em}@media (min-width:200ex){#index .two-column{column-count:2}}@media (min-width:300ex){#index .two-column{column-count:3}}dl{margin-bottom:2em}dl dl:last-child{margin-bottom:4em}dd{margin:0 0 1em 3em}#header-classes + dl > dd{margin-bottom:3em}dd dd{margin-left:2em}dd p{margin:10px 0}.name{background:#eee;font-weight:bold;font-size:.85em;padding:5px 10px;display:inline-block;min-width:40%}.name:hover{background:#e0e0e0}.name > span:first-child{white-space:nowrap}.name.class > span:nth-child(2){margin-left:.4em}.inherited{color:#999;border-left:5px solid #eee;padding-left:1em}.inheritance em{font-style:normal;font-weight:bold}.desc h2{font-weight:400;font-size:1.25em}.desc h3{font-size:1em}.desc dt code{background:inherit}.source summary,.git-link-div{color:#666;text-align:right;font-weight:400;font-size:.8em;text-transform:uppercase}.source summary > *{white-space:nowrap;cursor:pointer}.git-link{color:inherit;margin-left:1em}.source pre{max-height:500px;overflow:auto;margin:0}.source pre code{font-size:12px;overflow:visible}.hlist{list-style:none}.hlist li{display:inline}.hlist li:after{content:',\2002'}.hlist li:last-child:after{content:none}.hlist .hlist{display:inline;padding-left:1em}img{max-width:100%}.admonition{padding:.1em .5em;margin-bottom:1em}.admonition-title{font-weight:bold}.admonition.note,.admonition.info,.admonition.important{background:#aef}.admonition.todo,.admonition.versionadded,.admonition.tip,.admonition.hint{background:#dfd}.admonition.warning,.admonition.versionchanged,.admonition.deprecated{background:#fd4}.admonition.error,.admonition.danger,.admonition.caution{background:lightpink}</style>
<style media="screen and (min-width: 700px)">@media screen and (min-width:700px){#sidebar{width:30%}#content{width:70%;max-width:100ch;padding:3em 4em;border-left:1px solid #ddd}pre code{font-size:1em}.item .name{font-size:1em}main{display:flex;flex-direction:row-reverse;justify-content:flex-end}.toc ul ul,#index ul{padding-left:1.5em}.toc > ul > li{margin-top:.5em}}</style>
<style media="print">@media print{#sidebar h1{page-break-before:always}.source{display:none}}@media print{*{background:transparent !important;color:#000 !important;box-shadow:none !important;text-shadow:none !important}a[href]:after{content:" (" attr(href) ")";font-size:90%}a[href][title]:after{content:none}abbr[title]:after{content:" (" attr(title) ")"}.ir a:after,a[href^="javascript:"]:after,a[href^="#"]:after{content:""}pre,blockquote{border:1px solid #999;page-break-inside:avoid}thead{display:table-header-group}tr,img{page-break-inside:avoid}img{max-width:100% !important}@page{margin:0.5cm}p,h2,h3{orphans:3;widows:3}h1,h2,h3,h4,h5,h6{page-break-after:avoid}}</style>
</head>
<body>
<main>
<article id="content">
<header>
<h1 class="title">Module <code>dpm</code></h1>
</header>
<section id="section-intro">
<h1 id="weighted-dirichlet-process-gaussian-mixture-model">Weighted Dirichlet Process Gaussian Mixture Model</h1>
<p>This repo extends <a href="https://scikit-learn.org/stable/modules/generated/sklearn.mixture.BayesianGaussianMixture.html">sklearn.mixture.BayesianGaussianMixture</a> to support weighted training examples.</p>
<p>Complies with <code>sklearn.fit</code> API
-
<code>sample_weight</code> is a vector, same length as <code>X</code>. It must be <code>>=1</code> corresponding to duplicates or counts of observations otherwise the GMM model does not make sense.</p>
<p><a href="example.ipynb">example notebook</a></p>
<pre><code class="python">import numpy as np
from dpm.dpgmm import WeightedDPGMM
# make some data
num_clusters = 10
N = 200
x_means = 20 * np.random.rand(1, num_clusters, 2) - 10
y = np.random.randint(num_clusters, size=N)
x = .08 * np.random.randn(N, 1, 2)
temp = np.zeros((N, num_clusters, 1))
temp[np.arange(N), y, :] = 1
x = (x + x_means * temp).sum(1)
sample_weight = np.random.randint(1, 50, size=len(x))
# train the model
model = WeightedDPGMM(n_components=20, max_iter=1000, verbose=1)
yhat = model.fit_predict(x, sample_weight=sample_weight)
</code></pre>
<h3 id="true-vs-inferred-clusters">True vs Inferred Clusters</h3>
<p><img alt="True vs Inferred Clusters" src="imgs/sample.png"></p>
<h3 id="run-time-comparison-to-regular-unweighted-implementation">Run Time Comparison to Regular Unweighted Implementation</h3>
<p>Distribution over 20 trials. As we expect, each iteration is proportional to the number of samples. So if we bin the input then we get a speed up. More input points and larger bin sizes result in more gains (with loss of accuracy obvi). </p>
<p>This is with the following model parameters: <code>max_iter=1000, tol=1e-6, covariance_type="diag"</code></p>
<pre><code class="python">from timeit import timeit
tol = 1e-6
num_iters = 1000
cov_type = "diag"
def run_model(x,w,seed):
model = WeightedDPGMM(n_components=20, verbose=0, max_iter=num_iters, tol=tol, covariance_type=cov_type,random_state=seed)
labels = model.fit_predict(x, sample_weight=w)
def run_model_unweighted(x,seed):
model = BayesianGaussianMixture(n_components=20, verbose=0, max_iter=num_iters, tol=tol, covariance_type=cov_type,random_state=seed)
labels = model.fit_predict(x)
def time_model(x,w=None,kind="weighted",number = 1, seed = None):
if kind == "weighted":
dt = timeit(lambda: run_model(x,w,seed),number=number)
else:
dt = timeit(lambda: run_model_unweighted(x,seed),number=number)
return dict(dt = dt/number, kind=kind, size = len(x))
out = []
r = 0
seed=None
x_df = make_data()
for i in tqdm.trange(50):
for num_points in tqdm.tqdm(np.logspace(3,4,30),leave=False):
x_sample = x_df.sample(int(num_points))
x = x_sample.loc[:,["x","y"]].values
o = time_model(x,w=None,kind="unweighted",seed= seed)
o["og_size"] = int(num_points)
out.append(o)
x_df_rounded = x_sample.round(r).groupby(["x","y"]).size().to_frame("weight").reset_index()
x = x_df_rounded.loc[:,["x","y"]].values
w = x_df_rounded.loc[:,"weight"].values
o = time_model(x,w=w,kind="weighted",seed = seed)
o["og_size"] = int(num_points)
out.append(o)
</code></pre>
<p><a href="runtime_test.ipynb">run time test notebook</a></p>
<p><img alt="True vs Inferred Clusters" src="imgs/runtime.png"></p>
<details class="source">
<summary>
<span>Expand source code</span>
</summary>
<pre><code class="python">"""
.. include:: ../README.md
"""</code></pre>
</details>
</section>
<section>
<h2 class="section-title" id="header-submodules">Sub-modules</h2>
<dl>
<dt><code class="name"><a title="dpm.dpgmm" href="dpgmm.html">dpm.dpgmm</a></code></dt>
<dd>
<section class="desc"></section>
</dd>
</dl>
</section>
<section>
</section>
<section>
</section>
<section>
</section>
</article>
<nav id="sidebar">
<h1>Index</h1>
<div class="toc">
<ul>
<li><a href="#weighted-dirichlet-process-gaussian-mixture-model">Weighted Dirichlet Process Gaussian Mixture Model</a><ul>
<li><a href="#true-vs-inferred-clusters">True vs Inferred Clusters</a></li>
<li><a href="#run-time-comparison-to-regular-unweighted-implementation">Run Time Comparison to Regular Unweighted Implementation</a></li>
</ul>
</li>
</ul>
</div>
<ul id="index">
<li><h3><a href="#header-submodules">Sub-modules</a></h3>
<ul>
<li><code><a title="dpm.dpgmm" href="dpgmm.html">dpm.dpgmm</a></code></li>
</ul>
</li>
</ul>
</nav>
</main>
<footer id="footer">
<p>Generated by <a href="https://pdoc3.github.io/pdoc"><cite>pdoc</cite> 0.7.5</a>.</p>
</footer>
<script src="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/9.12.0/highlight.min.js"></script>
<script>hljs.initHighlightingOnLoad()</script>
</body>
</html>