sklearn.preprocessing.PolynomialFeatures (#269)

* adding supun's polyfeatures code * interaction_only flag breaks this, commenting out for now * removing extra dividers Co-authored-by: Matteo Interlandi <m.interlandi@gmail.com>
2020-08-30 13:52:03 -07:00 · 2020-08-30 13:52:03 -07:00 · e3cb0aedad
--- a/doc/html/hummingbird/ml/operator_converters/index.html
+++ b/doc/html/hummingbird/ml/operator_converters/index.html
@ -66,6 +66,7 @@ from .sklearn import nb as sklearn_nb  # noqa: E402
 from .sklearn import normalizer as sklearn_normalizer  # noqa: E402
 from .sklearn import one_hot_encoder as sklearn_ohe  # noqa: E402
 from .sklearn import pipeline  # noqa: E402
+from .sklearn import poly_features  # noqa: E402
 from .sklearn import scaler as sklearn_scaler  # noqa: E402
 from .sklearn import sv  # noqa: E402
 from . import lightgbm  # noqa: E402
--- a/doc/html/hummingbird/ml/operator_converters/sklearn/index.html
+++ b/doc/html/hummingbird/ml/operator_converters/sklearn/index.html
@ -90,6 +90,10 @@ All scikit-learn operators converters are stored under this package.
 <dd>
 <div class="desc"><p>Converters for operators necessary for supporting scikit-learn Pipelines.</p></div>
 </dd>
+<dt><code class="name"><a title="hummingbird.ml.operator_converters.sklearn.poly_features" href="poly_features.html">hummingbird.ml.operator_converters.sklearn.poly_features</a></code></dt>
+<dd>
+<div class="desc"><p>Converter for scikit-learn PolynomialFeatures.</p></div>
+</dd>
 <dt><code class="name"><a title="hummingbird.ml.operator_converters.sklearn.scaler" href="scaler.html">hummingbird.ml.operator_converters.sklearn.scaler</a></code></dt>
 <dd>
 <div class="desc"><p>Converters for scikit-learn scalers: RobustScaler, MaxAbsScaler, MinMaxScaler, StandardScaler.</p></div>
@ -137,6 +141,7 @@ All scikit-learn operators converters are stored under this package.
 <li><code><a title="hummingbird.ml.operator_converters.sklearn.normalizer" href="normalizer.html">hummingbird.ml.operator_converters.sklearn.normalizer</a></code></li>
 <li><code><a title="hummingbird.ml.operator_converters.sklearn.one_hot_encoder" href="one_hot_encoder.html">hummingbird.ml.operator_converters.sklearn.one_hot_encoder</a></code></li>
 <li><code><a title="hummingbird.ml.operator_converters.sklearn.pipeline" href="pipeline.html">hummingbird.ml.operator_converters.sklearn.pipeline</a></code></li>
+<li><code><a title="hummingbird.ml.operator_converters.sklearn.poly_features" href="poly_features.html">hummingbird.ml.operator_converters.sklearn.poly_features</a></code></li>
 <li><code><a title="hummingbird.ml.operator_converters.sklearn.scaler" href="scaler.html">hummingbird.ml.operator_converters.sklearn.scaler</a></code></li>
 <li><code><a title="hummingbird.ml.operator_converters.sklearn.sv" href="sv.html">hummingbird.ml.operator_converters.sklearn.sv</a></code></li>
 </ul>
--- a/doc/html/hummingbird/ml/operator_converters/sklearn/poly_features.html
+++ b/doc/html/hummingbird/ml/operator_converters/sklearn/poly_features.html
@ -0,0 +1,331 @@
+<!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 0.8.1" />
+<title>hummingbird.ml.operator_converters.sklearn.poly_features API documentation</title>
+<meta name="description" content="Converter for scikit-learn PolynomialFeatures." />
+<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}#sidebar > *:last-child{margin-bottom:2cm}.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{margin-top:.6em;font-weight:bold}@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%;height:100vh;overflow:auto;position:sticky;top:0}#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>
+<style>.homelink{display:block;font-size:2em;font-weight:bold;color:#555;padding-bottom:.5em;border-bottom:1px solid silver}.homelink:hover{color:inherit}.homelink img{max-width:20%;max-height:5em;margin:auto;margin-bottom:.3em}</style>
+<link rel="canonical" href="https://microsoft.github.io/hummingbird/ml/operator_converters/sklearn/poly_features.html">
+</head>
+<body>
+<main>
+<article id="content">
+<header>
+<h1 class="title">Module <code>hummingbird.ml.operator_converters.sklearn.poly_features</code></h1>
+</header>
+<section id="section-intro">
+<p>Converter for scikit-learn PolynomialFeatures.</p>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+<a href="https://github.com/microsoft/hummingbird/blob/master/hummingbird/ml/operator_converters/sklearn/poly_features.py#L0-L86" class="git-link">Browse git</a>
+</summary>
+<pre><code class="python"># -------------------------------------------------------------------------
+# Copyright (c) Microsoft Corporation. All rights reserved.
+# Licensed under the MIT License. See License.txt in the project root for
+# license information.
+# --------------------------------------------------------------------------
+
+&#34;&#34;&#34;
+Converter for scikit-learn PolynomialFeatures.
+&#34;&#34;&#34;
+from .._base_operator import BaseOperator
+from onnxconverter_common.registration import register_converter
+import torch
+
+
+class PolynomialFeatures(BaseOperator, torch.nn.Module):
+    &#34;&#34;&#34;
+    Class implementing PolynomialFeatures operators in PyTorch.
+
+    # TODO extend this class to support higher orders
+    &#34;&#34;&#34;
+
+    def __init__(self, n_features, degree, interaction_only, include_bias, device):
+        super(PolynomialFeatures, self).__init__()
+        self.transformer = True
+
+        self.n_features = n_features
+        self.interaction_only = interaction_only
+        self.include_bias = include_bias
+
+        indices = [i for j in range(n_features) for i in range(j * n_features + j, (j + 1) * n_features)]
+        self.n_poly_features = len(indices)
+        self.n_features = n_features
+        self.indices = torch.nn.Parameter(torch.LongTensor(indices), requires_grad=False)
+
+        self.bias = torch.nn.Parameter(torch.FloatTensor([1.0]), requires_grad=False)
+
+    def forward(self, x):
+        x_orig = x
+        x = x.view(-1, self.n_features, 1) * x.view(-1, 1, self.n_features)
+        x = x.view(-1, self.n_features ** 2)
+        x = torch.index_select(x, 1, self.indices)
+
+        # TODO: This gives mismatched elements
+        # if self.interaction_only:
+        #    if self.include_bias:
+        #        bias = self.bias.expand(x_orig.size()[0], 1)
+        #        return torch.cat([bias, x], dim=1)
+        #    else:
+        #        return x
+
+        if self.include_bias:
+            bias = self.bias.expand(x_orig.size()[0], 1)
+            return torch.cat([bias, x_orig, x], dim=1)
+        else:
+            return torch.cat([x_orig, x], dim=1)
+
+
+def convert_sklearn_poly_features(operator, device, extra_config):
+    &#34;&#34;&#34;
+    Converter for `sklearn.preprocessing.PolynomialFeatures`
+
+    Currently this supports only degree 2, and does not support interaction_only
+
+    Args:
+        operator: An operator wrapping a `sklearn.preprocessing.PolynomialFeatures` model
+        device: String defining the type of device the converted operator should be run on
+        extra_config: Extra configuration used to select the best conversion strategy
+
+    Returns:
+        A PyTorch model
+    &#34;&#34;&#34;
+
+    if operator.raw_operator.interaction_only:
+        raise NotImplementedError(&#34;Hummingbird does not currently support interaction_only flag for PolynomialFeatures&#34;)
+
+    if operator.raw_operator.degree != 2:
+        raise NotImplementedError(&#34;Hummingbird currently only supports degree 2 for PolynomialFeatures&#34;)
+    return PolynomialFeatures(
+        operator.raw_operator.n_input_features_,
+        operator.raw_operator.degree,
+        operator.raw_operator.interaction_only,
+        operator.raw_operator.include_bias,
+        device,
+    )
+
+
+register_converter(&#34;SklearnPolynomialFeatures&#34;, convert_sklearn_poly_features)</code></pre>
+</details>
+</section>
+<section>
+</section>
+<section>
+</section>
+<section>
+<h2 class="section-title" id="header-functions">Functions</h2>
+<dl>
+<dt id="hummingbird.ml.operator_converters.sklearn.poly_features.convert_sklearn_poly_features"><code class="name flex">
+<span>def <span class="ident">convert_sklearn_poly_features</span></span>(<span>operator, device, extra_config)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Converter for <code>sklearn.preprocessing.PolynomialFeatures</code></p>
+<p>Currently this supports only degree 2, and does not support interaction_only</p>
+<h2 id="args">Args</h2>
+<dl>
+<dt><strong><code>operator</code></strong></dt>
+<dd>An operator wrapping a <code>sklearn.preprocessing.PolynomialFeatures</code> model</dd>
+<dt><strong><code>device</code></strong></dt>
+<dd>String defining the type of device the converted operator should be run on</dd>
+<dt><strong><code>extra_config</code></strong></dt>
+<dd>Extra configuration used to select the best conversion strategy</dd>
+</dl>
+<h2 id="returns">Returns</h2>
+<dl>
+<dt><code>A PyTorch model</code></dt>
+<dd>&nbsp;</dd>
+</dl></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+<a href="https://github.com/microsoft/hummingbird/blob/master/hummingbird/ml/operator_converters/sklearn/poly_features.py#L58-L84" class="git-link">Browse git</a>
+</summary>
+<pre><code class="python">def convert_sklearn_poly_features(operator, device, extra_config):
+    &#34;&#34;&#34;
+    Converter for `sklearn.preprocessing.PolynomialFeatures`
+
+    Currently this supports only degree 2, and does not support interaction_only
+
+    Args:
+        operator: An operator wrapping a `sklearn.preprocessing.PolynomialFeatures` model
+        device: String defining the type of device the converted operator should be run on
+        extra_config: Extra configuration used to select the best conversion strategy
+
+    Returns:
+        A PyTorch model
+    &#34;&#34;&#34;
+
+    if operator.raw_operator.interaction_only:
+        raise NotImplementedError(&#34;Hummingbird does not currently support interaction_only flag for PolynomialFeatures&#34;)
+
+    if operator.raw_operator.degree != 2:
+        raise NotImplementedError(&#34;Hummingbird currently only supports degree 2 for PolynomialFeatures&#34;)
+    return PolynomialFeatures(
+        operator.raw_operator.n_input_features_,
+        operator.raw_operator.degree,
+        operator.raw_operator.interaction_only,
+        operator.raw_operator.include_bias,
+        device,
+    )</code></pre>
+</details>
+</dd>
+</dl>
+</section>
+<section>
+<h2 class="section-title" id="header-classes">Classes</h2>
+<dl>
+<dt id="hummingbird.ml.operator_converters.sklearn.poly_features.PolynomialFeatures"><code class="flex name class">
+<span>class <span class="ident">PolynomialFeatures</span></span>
+<span>(</span><span>n_features, degree, interaction_only, include_bias, device)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Class implementing PolynomialFeatures operators in PyTorch.</p>
+<h1 id="todo-extend-this-class-to-support-higher-orders">TODO extend this class to support higher orders</h1></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+<a href="https://github.com/microsoft/hummingbird/blob/master/hummingbird/ml/operator_converters/sklearn/poly_features.py#L15-L55" class="git-link">Browse git</a>
+</summary>
+<pre><code class="python">class PolynomialFeatures(BaseOperator, torch.nn.Module):
+    &#34;&#34;&#34;
+    Class implementing PolynomialFeatures operators in PyTorch.
+
+    # TODO extend this class to support higher orders
+    &#34;&#34;&#34;
+
+    def __init__(self, n_features, degree, interaction_only, include_bias, device):
+        super(PolynomialFeatures, self).__init__()
+        self.transformer = True
+
+        self.n_features = n_features
+        self.interaction_only = interaction_only
+        self.include_bias = include_bias
+
+        indices = [i for j in range(n_features) for i in range(j * n_features + j, (j + 1) * n_features)]
+        self.n_poly_features = len(indices)
+        self.n_features = n_features
+        self.indices = torch.nn.Parameter(torch.LongTensor(indices), requires_grad=False)
+
+        self.bias = torch.nn.Parameter(torch.FloatTensor([1.0]), requires_grad=False)
+
+    def forward(self, x):
+        x_orig = x
+        x = x.view(-1, self.n_features, 1) * x.view(-1, 1, self.n_features)
+        x = x.view(-1, self.n_features ** 2)
+        x = torch.index_select(x, 1, self.indices)
+
+        # TODO: This gives mismatched elements
+        # if self.interaction_only:
+        #    if self.include_bias:
+        #        bias = self.bias.expand(x_orig.size()[0], 1)
+        #        return torch.cat([bias, x], dim=1)
+        #    else:
+        #        return x
+
+        if self.include_bias:
+            bias = self.bias.expand(x_orig.size()[0], 1)
+            return torch.cat([bias, x_orig, x], dim=1)
+        else:
+            return torch.cat([x_orig, x], dim=1)</code></pre>
+</details>
+<h3>Ancestors</h3>
+<ul class="hlist">
+<li>hummingbird.ml.operator_converters._base_operator.BaseOperator</li>
+<li>abc.ABC</li>
+<li>torch.nn.modules.module.Module</li>
+</ul>
+<h3>Methods</h3>
+<dl>
+<dt id="hummingbird.ml.operator_converters.sklearn.poly_features.PolynomialFeatures.forward"><code class="name flex">
+<span>def <span class="ident">forward</span></span>(<span>self, x)</span>
+</code></dt>
+<dd>
+<div class="desc"><p>Defines the computation performed at every call.</p>
+<p>Should be overridden by all subclasses.</p>
+<div class="admonition note">
+<p class="admonition-title">Note</p>
+<p>Although the recipe for forward pass needs to be defined within
+this function, one should call the :class:<code>Module</code> instance afterwards
+instead of this since the former takes care of running the
+registered hooks while the latter silently ignores them.</p>
+</div></div>
+<details class="source">
+<summary>
+<span>Expand source code</span>
+<a href="https://github.com/microsoft/hummingbird/blob/master/hummingbird/ml/operator_converters/sklearn/poly_features.py#L37-L55" class="git-link">Browse git</a>
+</summary>
+<pre><code class="python">def forward(self, x):
+    x_orig = x
+    x = x.view(-1, self.n_features, 1) * x.view(-1, 1, self.n_features)
+    x = x.view(-1, self.n_features ** 2)
+    x = torch.index_select(x, 1, self.indices)
+
+    # TODO: This gives mismatched elements
+    # if self.interaction_only:
+    #    if self.include_bias:
+    #        bias = self.bias.expand(x_orig.size()[0], 1)
+    #        return torch.cat([bias, x], dim=1)
+    #    else:
+    #        return x
+
+    if self.include_bias:
+        bias = self.bias.expand(x_orig.size()[0], 1)
+        return torch.cat([bias, x_orig, x], dim=1)
+    else:
+        return torch.cat([x_orig, x], dim=1)</code></pre>
+</details>
+</dd>
+</dl>
+</dd>
+</dl>
+</section>
+</article>
+<nav id="sidebar">
+<header>
+<a class="homelink" rel="home" title="Hummingbird Home" href="https://github.com/microsoft/hummingbird"> Hummingbird
+</a>
+</header>
+<h1>Index</h1>
+<div class="toc">
+<ul></ul>
+</div>
+<ul id="index">
+<li><h3>Super-module</h3>
+<ul>
+<li><code><a title="hummingbird.ml.operator_converters.sklearn" href="index.html">hummingbird.ml.operator_converters.sklearn</a></code></li>
+</ul>
+</li>
+<li><h3><a href="#header-functions">Functions</a></h3>
+<ul class="">
+<li><code><a title="hummingbird.ml.operator_converters.sklearn.poly_features.convert_sklearn_poly_features" href="#hummingbird.ml.operator_converters.sklearn.poly_features.convert_sklearn_poly_features">convert_sklearn_poly_features</a></code></li>
+</ul>
+</li>
+<li><h3><a href="#header-classes">Classes</a></h3>
+<ul>
+<li>
+<h4><code><a title="hummingbird.ml.operator_converters.sklearn.poly_features.PolynomialFeatures" href="#hummingbird.ml.operator_converters.sklearn.poly_features.PolynomialFeatures">PolynomialFeatures</a></code></h4>
+<ul class="">
+<li><code><a title="hummingbird.ml.operator_converters.sklearn.poly_features.PolynomialFeatures.forward" href="#hummingbird.ml.operator_converters.sklearn.poly_features.PolynomialFeatures.forward">forward</a></code></li>
+</ul>
+</li>
+</ul>
+</li>
+</ul>
+</nav>
+</main>
+<footer id="footer">
+<p>Generated by <a href="https://pdoc3.github.io/pdoc"><cite>pdoc</cite> 0.8.1</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>
--- a/doc/html/hummingbird/ml/supported.html
+++ b/doc/html/hummingbird/ml/supported.html
@ -51,6 +51,7 @@ MLPClassifier,
 MultinomialNB,
 Normalizer,
 OneHotEncoder,
+PolynomialFeatures,
 RandomForestClassifier,
 RandomForestRegressor,
 RobustScaler,
@ -111,6 +112,7 @@ MLPClassifier,
 MultinomialNB,
 Normalizer,
 OneHotEncoder,
+PolynomialFeatures,
 RandomForestClassifier,
 RandomForestRegressor,
 RobustScaler,
@ -188,6 +190,7 @@ def _build_sklearn_operator_list():
            MinMaxScaler,
            Normalizer,
            OneHotEncoder,
+            PolynomialFeatures,
            RobustScaler,
            StandardScaler,
        )
@ -231,6 +234,7 @@ def _build_sklearn_operator_list():
            MaxAbsScaler,
            MinMaxScaler,
            Normalizer,
+            PolynomialFeatures,
            RobustScaler,
            StandardScaler,
            # Feature selection
--- a/hummingbird/ml/operator_converters/init.py
+++ b/hummingbird/ml/operator_converters/init.py
@ -36,6 +36,7 @@ from .sklearn import nb as sklearn_nb  # noqa: E402
 from .sklearn import normalizer as sklearn_normalizer  # noqa: E402
 from .sklearn import one_hot_encoder as sklearn_ohe  # noqa: E402
 from .sklearn import pipeline  # noqa: E402
+from .sklearn import poly_features  # noqa: E402
 from .sklearn import scaler as sklearn_scaler  # noqa: E402
 from .sklearn import sv  # noqa: E402
 from . import lightgbm  # noqa: E402
--- a/hummingbird/ml/operator_converters/sklearn/poly_features.py
+++ b/hummingbird/ml/operator_converters/sklearn/poly_features.py
@ -0,0 +1,87 @@
+# -------------------------------------------------------------------------
+# Copyright (c) Microsoft Corporation. All rights reserved.
+# Licensed under the MIT License. See License.txt in the project root for
+# license information.
+# --------------------------------------------------------------------------
+
+"""
+Converter for scikit-learn PolynomialFeatures.
+"""
+from .._base_operator import BaseOperator
+from onnxconverter_common.registration import register_converter
+import torch
+
+
+class PolynomialFeatures(BaseOperator, torch.nn.Module):
+    """
+    Class implementing PolynomialFeatures operators in PyTorch.
+
+    # TODO extend this class to support higher orders
+    """
+
+    def __init__(self, n_features, degree, interaction_only, include_bias, device):
+        super(PolynomialFeatures, self).__init__()
+        self.transformer = True
+
+        self.n_features = n_features
+        self.interaction_only = interaction_only
+        self.include_bias = include_bias
+
+        indices = [i for j in range(n_features) for i in range(j * n_features + j, (j + 1) * n_features)]
+        self.n_poly_features = len(indices)
+        self.n_features = n_features
+        self.indices = torch.nn.Parameter(torch.LongTensor(indices), requires_grad=False)
+
+        self.bias = torch.nn.Parameter(torch.FloatTensor([1.0]), requires_grad=False)
+
+    def forward(self, x):
+        x_orig = x
+        x = x.view(-1, self.n_features, 1) * x.view(-1, 1, self.n_features)
+        x = x.view(-1, self.n_features ** 2)
+        x = torch.index_select(x, 1, self.indices)
+
+        # TODO: This gives mismatched elements
+        # if self.interaction_only:
+        #    if self.include_bias:
+        #        bias = self.bias.expand(x_orig.size()[0], 1)
+        #        return torch.cat([bias, x], dim=1)
+        #    else:
+        #        return x
+
+        if self.include_bias:
+            bias = self.bias.expand(x_orig.size()[0], 1)
+            return torch.cat([bias, x_orig, x], dim=1)
+        else:
+            return torch.cat([x_orig, x], dim=1)
+
+
+def convert_sklearn_poly_features(operator, device, extra_config):
+    """
+    Converter for `sklearn.preprocessing.PolynomialFeatures`
+
+    Currently this supports only degree 2, and does not support interaction_only
+
+    Args:
+        operator: An operator wrapping a `sklearn.preprocessing.PolynomialFeatures` model
+        device: String defining the type of device the converted operator should be run on
+        extra_config: Extra configuration used to select the best conversion strategy
+
+    Returns:
+        A PyTorch model
+    """
+
+    if operator.raw_operator.interaction_only:
+        raise NotImplementedError("Hummingbird does not currently support interaction_only flag for PolynomialFeatures")
+
+    if operator.raw_operator.degree != 2:
+        raise NotImplementedError("Hummingbird currently only supports degree 2 for PolynomialFeatures")
+    return PolynomialFeatures(
+        operator.raw_operator.n_input_features_,
+        operator.raw_operator.degree,
+        operator.raw_operator.interaction_only,
+        operator.raw_operator.include_bias,
+        device,
+    )
+
+
+register_converter("SklearnPolynomialFeatures", convert_sklearn_poly_features)
--- a/hummingbird/ml/supported.py
+++ b/hummingbird/ml/supported.py
@ -36,6 +36,7 @@ MLPClassifier,
 MultinomialNB,
 Normalizer,
 OneHotEncoder,
+PolynomialFeatures,
 RandomForestClassifier,
 RandomForestRegressor,
 RobustScaler,
@ -113,6 +114,7 @@ def _build_sklearn_operator_list():
            MinMaxScaler,
            Normalizer,
            OneHotEncoder,
+            PolynomialFeatures,
            RobustScaler,
            StandardScaler,
        )
@ -156,6 +158,7 @@ def _build_sklearn_operator_list():
            MaxAbsScaler,
            MinMaxScaler,
            Normalizer,
+            PolynomialFeatures,
            RobustScaler,
            StandardScaler,
            # Feature selection
--- a/tests/test_sklearn_poly_features_converter.py
+++ b/tests/test_sklearn_poly_features_converter.py
@ -0,0 +1,59 @@
+"""
+Tests sklearn Binarizer converter
+"""
+import unittest
+import warnings
+
+import numpy as np
+import torch
+from sklearn.preprocessing import PolynomialFeatures
+
+import hummingbird.ml
+
+
+class TestSklearnPolynomialFeatures(unittest.TestCase):
+    def _test_sklearn_polynomial_featurizer(self, data, model):
+
+        data_tensor = torch.from_numpy(data)
+
+        torch_model = hummingbird.ml.convert(model, "torch")
+        self.assertIsNotNone(torch_model)
+        np.testing.assert_allclose(
+            model.transform(data), torch_model.transform(data_tensor), rtol=1e-06, atol=1e-06,
+        )
+
+    def test_sklearn_poly_feat_with_bias(self):
+        data = np.array([[1.2, 3.2, 1.3, -5.6], [4.3, -3.2, 5.7, 1.0], [0, 3.2, 4.7, -8.9]], dtype=np.float32)
+        model = PolynomialFeatures(degree=2, include_bias=True, order="F").fit(data)
+        self._test_sklearn_polynomial_featurizer(data, model)
+
+    def test_sklearn_poly_feat_with_no_bias(self):
+        data = np.array([[1.2, 3.2, 1.3, -5.6], [4.3, -3.2, 5.7, 1.0], [0, 3.2, 4.7, -8.9]], dtype=np.float32)
+        model = PolynomialFeatures(degree=2, include_bias=False, order="F").fit(data)
+        self._test_sklearn_polynomial_featurizer(data, model)
+
+    # TODO: interaction is not currently supported (bug)
+    # def test_sklearn_poly_feat_with_interaction_and_bias(self):
+    #    data = np.array([[1.2, 3.2, 1.3, -5.6], [4.3, -3.2, 5.7, 1.0], [0, 3.2, 4.7, -8.9]], dtype=np.float32)
+    #    model = PolynomialFeatures(degree=2, include_bias=True, order="F", interaction_only=True).fit(data)
+    #    self._test_sklearn_polynomial_featurizer(data, model)
+
+    # def test_sklearn_poly_feat_with_interaction_and_no_bias(self):
+    #    data = np.array([[1.2, 3.2, 1.3, -5.6], [4.3, -3.2, 5.7, 1.0], [0, 3.2, 4.7, -8.9]], dtype=np.float32)
+    #    model = PolynomialFeatures(degree=2, include_bias=False, order="F", interaction_only=True).fit(data)
+    #    self._test_sklearn_polynomial_featurizer(data, model)
+
+    def test_sklearn_poly_featurizer_raises(self):
+        data = np.array([[1.2, 3.2, 1.3, -5.6], [4.3, -3.2, 5.7, 1.0], [0, 3.2, 4.7, -8.9]], dtype=np.float32)
+
+        # TODO: delete when implemented
+        model = PolynomialFeatures(degree=4, include_bias=True, order="F").fit(data)
+        self.assertRaises(NotImplementedError, hummingbird.ml.convert, model, "torch")
+
+        # TODO: delete when implemented
+        model = PolynomialFeatures(degree=2, interaction_only=True, order="F").fit(data)
+        self.assertRaises(NotImplementedError, hummingbird.ml.convert, model, "torch")
+
+
+if __name__ == "__main__":
+    unittest.main()