update changelog + docs

Author: ardunn

CHANGELOG.md

@@ -2,7 +2,24 @@
 
 ## [Unreleased](https://github.com/hackingmaterials/automatminer/tree/HEAD)
 
-[Full Changelog](https://github.com/hackingmaterials/automatminer/compare/v2019.01.25_beta...HEAD)
+[Full Changelog](https://github.com/hackingmaterials/automatminer/compare/v2019.01.26_beta...HEAD)
+
+**Closed issues:**
+
+- Nose ---\> unittest  [\#171](https://github.com/hackingmaterials/automatminer/issues/171)
+- Fix benchmarking [\#170](https://github.com/hackingmaterials/automatminer/issues/170)
+- Should add to PyPi [\#168](https://github.com/hackingmaterials/automatminer/issues/168)
+- An adapter to run a single model [\#165](https://github.com/hackingmaterials/automatminer/issues/165)
+- Add option to remove specific features [\#159](https://github.com/hackingmaterials/automatminer/issues/159)
+- Analytics module needs tests [\#133](https://github.com/hackingmaterials/automatminer/issues/133)
+
+**Merged pull requests:**
+
+- Update codacy and circleCI configs [\#173](https://github.com/hackingmaterials/automatminer/pull/173) ([utf](https://github.com/utf))
+- Add optional to manually keep/remove features [\#172](https://github.com/hackingmaterials/automatminer/pull/172) ([utf](https://github.com/utf))
+
+## [v2019.01.26_beta](https://github.com/hackingmaterials/automatminer/tree/v2019.01.26_beta) (2019-01-26)
+[Full Changelog](https://github.com/hackingmaterials/automatminer/compare/v2019.01.25_beta...v2019.01.26_beta)
 
 **Closed issues:**
 

automatminer/__init__.py

@@ -7,4 +7,4 @@
 __author__ = 'Alex Dunn, Qi Wang, Alex Ganose, Daniel Dopp, Anubhav Jain'
 __author_email__ = '[email protected]'
 __license__ = 'Modified BSD'
-__version__ = '2019.01.26_beta'
+__version__ = '2019.02.02_beta'

docs/_sources/index.rst.txt

@@ -9,16 +9,21 @@
 .. image:: _static/logo_med.png
    :alt: server
    :align: center
-   :width: 400px
+   :width: 600px
 
 
-Automatminer is a tool for automatically creating complete machine learning pipelines for materials science, which includes automatic featurization with `matminer <https://github.com/hackingmaterials/matminer>`_, feature reduction, and an AutoML backend. Put in a materials dataset, get out a machine that predicts materials properties.
+Automatminer is a tool for *automatically* creating **complete** machine learning pipelines for materials science, including automatic featurization with `matminer <https://github.com/hackingmaterials/matminer>`_, feature reduction, and an AutoML backend. Put in a materials dataset, get out a machine that predicts materials properties.
 
+How it works
+=============
+
+Automatminer automatically decorates a dataset using hundreds of descriptor techniques from matminer's descriptor library, picks the most useful features for learning, and runs a separate AutoML pipeline using TPOT. Once a pipeline has been fit, it can be examined with skater's interpretability tools, summarized in a text file, saved to disk, or used to make new predictions.
 
 .. image:: _static/automatminer_big.jpg
    :alt: server
    :align: center
 
+
 Here's an example of training on known data, and extending the model to out of sample data.
 
 .. code-block:: python
@@ -33,7 +38,7 @@ Here's an example of training on known data, and extending the model to out of s
     predicted_df = pipe.predict(unknown_df, "band gap")
 
 
-Alternatively, run a nested cross validation benchmark on a known dataset, and then compare the results against your own ML models:
+Or, run a (relatively) rigorous nested cross validation benchmark on a known dataset, and then compare the results against your own ML models:
 
 .. code-block:: python
 
@@ -45,58 +50,40 @@ Alternatively, run a nested cross validation benchmark on a known dataset, and t
 
 
 
-automatminer is applicable to many problems
--------------------------------------------
+automatminer is widely applicable
+===========================================
 
 Automatminer can work with many kinds of data:
-*   both computational and experimental data
-*   small (~100 samples) to moderate (~100k samples) sized datasets
-*   crystalline datasets
-*   composition-only (i.e., unknown phases) datasets
-*   datasets containing electronic bandstructures or density of states
-
-...Many kinds of target properties
-*   electronic
-mechanical
-thermodynamic
-any other kind of property
-
-...And many featurization (descriptor) techniques:
-*list them*
-
-Automatminer automatically decorates a dataset using hundreds of descriptor techniques from matminer's descriptor library, picks the most useful features for learning, and runs a separate AutoML pipeline using TPOT. Once a pipeline has been fit, it can be examined with skater's interpretability tools, summarized in a text file, saved to disk, or used to make new predictions.
-
-
-Code Examples
-=============
+----------------------------------------------
+-   both computational and experimental data
+-   small (~100 samples) to moderate (~100k samples) sized datasets
+-   crystalline datasets
+-   composition-only (i.e., unknown phases) datasets
+-   datasets containing electronic bandstructures or density of states
 
-The easiest (and most automatic) way to use automatminer is through the MatPipe object. First, fit the MatPipe to a dataframe containing materials objects such as chemical compositions (or pymatgen Structures) and some material target property.
-```python
+Many kinds of target properties:
+--------------------------------
+-   electronic
+-   mechanical
+-   thermodynamic
+-   any other kind of property
 
-```
+And many featurization (descriptor) techniques:
+-----------------------------------------------
+See `matminer's Table of Featurizers <https://hackingmaterials.github.io/matminer/featurizer_summary.html>`_ for a full (and growing) list.
 
-Now use your pipeline to predict the properties of some other data, such as a new composition or structure.
-```python
 
-```
 
-You can also use it to benchmark against other machine learning models with the `benchmark` method of MatPipe, which runs a Nested Cross Validation. The Nested CV scheme
-is typically a more robust way of estimating an ML pipeline's generalizaiton error than a simple train/validation/test split.
-```python
-from automatminer.pipeline import MatPipe
-from sklearn.model_selection import KFold
-
-pipe = MatPipe()
-predictions_per_fold = pipe.benchmark(df, "bulk modulus", KFold(n_splits=5))
-```
-
-Once a MatPipe has been fit, you can examine it internally to see how it works using `pipe.digest()`; or pickle it for later with `pipe.save()`.
+Full Code Examples
+==================
 
-### Citing automatminer
-We are in the process of writing a paper for automatminer. In the meantime, please use the citation given in the matminer repo.
+Citing automatminer
+===================
+We are in the process of writing a paper for automatminer. In the meantime, please use the citation given in the `matminer repo <https://github.com/hackingmaterials/matminer>`_.
 
-## Contributing
-Interested in contributing? See our [contribution guidelines](https://github.com/hackingmaterials/automatminer/blob/master/CONTRIBUTING.md) and make a pull request! Please submit questions, issues / bug reports, and all other communication through the [matminer Google Group](https://groups.google.com/forum/#!forum/matminer).
+Contributing
+============
+Interested in contributing? See our `contribution guidelines <https://github.com/hackingmaterials/automatminer/blob/master/CONTRIBUTING.md>`_ and make a pull request! Please submit questions, issues / bug reports, and all other communication through the  `matminer Google Group <https://groups.google.com/forum/#!forum/matminer>`_.
 
 
 

docs/index.html

@@ -6,7 +6,7 @@
   <head>
     <meta http-equiv="X-UA-Compatible" content="IE=Edge" />
     <meta http-equiv="Content-Type" content="text/html; charset=utf-8" />
-    <title>automatminer is applicable to many problems &#8212; automatminer 2019.01.26_beta documentation</title>
+    <title>How it works &#8212; automatminer 2019.01.26_beta documentation</title>
     <link rel="stylesheet" href="_static/alabaster.css" type="text/css" />
     <link rel="stylesheet" href="_static/pygments.css" type="text/css" />
     <script type="text/javascript" id="documentation_options" data-url_root="./" src="_static/documentation_options.js"></script>
@@ -32,8 +32,11 @@
           <div class="body" role="main">
 
   <style> .red {color:#aa0060; font-weight:bold; font-size:16px} </style><p><span class="red">WARNING! These docs are incomplete. Read and use at your own risk!`</span></p>
-<a class="reference internal image-reference" href="_images/logo_med.png"><img alt="server" class="align-center" src="_images/logo_med.png" style="width: 400px;" /></a>
-<p>Automatminer is a tool for automatically creating complete machine learning pipelines for materials science, which includes automatic featurization with <a class="reference external" href="https://github.com/hackingmaterials/matminer">matminer</a>, feature reduction, and an AutoML backend. Put in a materials dataset, get out a machine that predicts materials properties.</p>
+<a class="reference internal image-reference" href="_images/logo_med.png"><img alt="server" class="align-center" src="_images/logo_med.png" style="width: 600px;" /></a>
+<p>Automatminer is a tool for <em>automatically</em> creating <strong>complete</strong> machine learning pipelines for materials science, including automatic featurization with <a class="reference external" href="https://github.com/hackingmaterials/matminer">matminer</a>, feature reduction, and an AutoML backend. Put in a materials dataset, get out a machine that predicts materials properties.</p>
+<div class="section" id="how-it-works">
+<h1>How it works<a class="headerlink" href="#how-it-works" title="Permalink to this headline">¶</a></h1>
+<p>Automatminer automatically decorates a dataset using hundreds of descriptor techniques from matminer’s descriptor library, picks the most useful features for learning, and runs a separate AutoML pipeline using TPOT. Once a pipeline has been fit, it can be examined with skater’s interpretability tools, summarized in a text file, saved to disk, or used to make new predictions.</p>
 <img alt="server" class="align-center" src="_images/automatminer_big.jpg" />
 <p>Here’s an example of training on known data, and extending the model to out of sample data.</p>
 <div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="kn">from</span> <span class="nn">automatminer.pipeline</span> <span class="kn">import</span> <span class="n">MatPipe</span>
@@ -46,60 +49,59 @@
 <span class="n">predicted_df</span> <span class="o">=</span> <span class="n">pipe</span><span class="o">.</span><span class="n">predict</span><span class="p">(</span><span class="n">unknown_df</span><span class="p">,</span> <span class="s2">&quot;band gap&quot;</span><span class="p">)</span>
 </pre></div>
 </div>
-<p>Alternatively, run a nested cross validation benchmark on a known dataset, and then compare the results against your own ML models:</p>
+<p>Or, run a (relatively) rigorous nested cross validation benchmark on a known dataset, and then compare the results against your own ML models:</p>
 <div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="kn">from</span> <span class="nn">automatminer.pipeline</span> <span class="kn">import</span> <span class="n">MatPipe</span>
 <span class="kn">from</span> <span class="nn">sklearn.model_selection</span> <span class="kn">import</span> <span class="n">KFold</span>
 
 <span class="n">pipe</span> <span class="o">=</span> <span class="n">MatPipe</span><span class="p">()</span>
 <span class="n">predictions_per_fold</span> <span class="o">=</span> <span class="n">pipe</span><span class="o">.</span><span class="n">benchmark</span><span class="p">(</span><span class="n">df</span><span class="p">,</span> <span class="s2">&quot;bulk modulus&quot;</span><span class="p">,</span> <span class="n">KFold</span><span class="p">(</span><span class="n">n_splits</span><span class="o">=</span><span class="mi">5</span><span class="p">))</span>
 </pre></div>
 </div>
-<div class="section" id="automatminer-is-applicable-to-many-problems">
-<h1>automatminer is applicable to many problems<a class="headerlink" href="#automatminer-is-applicable-to-many-problems" title="Permalink to this headline">¶</a></h1>
-<p>Automatminer can work with many kinds of data:
-*   both computational and experimental data
-*   small (~100 samples) to moderate (~100k samples) sized datasets
-*   crystalline datasets
-*   composition-only (i.e., unknown phases) datasets
-*   datasets containing electronic bandstructures or density of states</p>
-<p>…Many kinds of target properties
-*   electronic
-mechanical
-thermodynamic
-any other kind of property</p>
-<p>…And many featurization (descriptor) techniques:
-<em>list them</em></p>
-<p>Automatminer automatically decorates a dataset using hundreds of descriptor techniques from matminer’s descriptor library, picks the most useful features for learning, and runs a separate AutoML pipeline using TPOT. Once a pipeline has been fit, it can be examined with skater’s interpretability tools, summarized in a text file, saved to disk, or used to make new predictions.</p>
-<div class="section" id="code-examples">
-<h2>Code Examples<a class="headerlink" href="#code-examples" title="Permalink to this headline">¶</a></h2>
-<p>The easiest (and most automatic) way to use automatminer is through the MatPipe object. First, fit the MatPipe to a dataframe containing materials objects such as chemical compositions (or pymatgen Structures) and some material target property.
-<a href="#id1"><span class="problematic" id="id2">``</span></a><a href="#id3"><span class="problematic" id="id4">`</span></a>python</p>
-<p><a href="#id5"><span class="problematic" id="id6">``</span></a><a href="#id7"><span class="problematic" id="id8">`</span></a></p>
-<p>Now use your pipeline to predict the properties of some other data, such as a new composition or structure.
-<a href="#id9"><span class="problematic" id="id10">``</span></a><a href="#id11"><span class="problematic" id="id12">`</span></a>python</p>
-<p><a href="#id13"><span class="problematic" id="id14">``</span></a><a href="#id15"><span class="problematic" id="id16">`</span></a></p>
-<p>You can also use it to benchmark against other machine learning models with the <cite>benchmark</cite> method of MatPipe, which runs a Nested Cross Validation. The Nested CV scheme
-is typically a more robust way of estimating an ML pipeline’s generalizaiton error than a simple train/validation/test split.
-<a href="#id17"><span class="problematic" id="id18">``</span></a><a href="#id19"><span class="problematic" id="id20">`</span></a>python
-from automatminer.pipeline import MatPipe
-from sklearn.model_selection import KFold</p>
-<p>pipe = MatPipe()
-predictions_per_fold = pipe.benchmark(df, “bulk modulus”, KFold(n_splits=5))
-<a href="#id21"><span class="problematic" id="id22">``</span></a><a href="#id23"><span class="problematic" id="id24">`</span></a></p>
-<p>Once a MatPipe has been fit, you can examine it internally to see how it works using <cite>pipe.digest()</cite>; or pickle it for later with <cite>pipe.save()</cite>.</p>
-<p>### Citing automatminer
-We are in the process of writing a paper for automatminer. In the meantime, please use the citation given in the matminer repo.</p>
-<p>## Contributing
-Interested in contributing? See our [contribution guidelines](<a class="reference external" href="https://github.com/hackingmaterials/automatminer/blob/master/CONTRIBUTING.md">https://github.com/hackingmaterials/automatminer/blob/master/CONTRIBUTING.md</a>) and make a pull request! Please submit questions, issues / bug reports, and all other communication through the [matminer Google Group](<a class="reference external" href="https://groups.google.com/forum/#!forum/matminer">https://groups.google.com/forum/#!forum/matminer</a>).</p>
+</div>
+<div class="section" id="automatminer-is-widely-applicable">
+<h1>automatminer is widely applicable<a class="headerlink" href="#automatminer-is-widely-applicable" title="Permalink to this headline">¶</a></h1>
+<div class="section" id="automatminer-can-work-with-many-kinds-of-data">
+<h2>Automatminer can work with many kinds of data:<a class="headerlink" href="#automatminer-can-work-with-many-kinds-of-data" title="Permalink to this headline">¶</a></h2>
+<ul class="simple">
+<li>both computational and experimental data</li>
+<li>small (~100 samples) to moderate (~100k samples) sized datasets</li>
+<li>crystalline datasets</li>
+<li>composition-only (i.e., unknown phases) datasets</li>
+<li>datasets containing electronic bandstructures or density of states</li>
+</ul>
+</div>
+<div class="section" id="many-kinds-of-target-properties">
+<h2>Many kinds of target properties:<a class="headerlink" href="#many-kinds-of-target-properties" title="Permalink to this headline">¶</a></h2>
+<ul class="simple">
+<li>electronic</li>
+<li>mechanical</li>
+<li>thermodynamic</li>
+<li>any other kind of property</li>
+</ul>
+</div>
+<div class="section" id="and-many-featurization-descriptor-techniques">
+<h2>And many featurization (descriptor) techniques:<a class="headerlink" href="#and-many-featurization-descriptor-techniques" title="Permalink to this headline">¶</a></h2>
+<p>See <a class="reference external" href="https://hackingmaterials.github.io/matminer/featurizer_summary.html">matminer’s Table of Featurizers</a> for a full (and growing) list.</p>
+</div>
+</div>
+<div class="section" id="full-code-examples">
+<h1>Full Code Examples<a class="headerlink" href="#full-code-examples" title="Permalink to this headline">¶</a></h1>
+</div>
+<div class="section" id="citing-automatminer">
+<h1>Citing automatminer<a class="headerlink" href="#citing-automatminer" title="Permalink to this headline">¶</a></h1>
+<p>We are in the process of writing a paper for automatminer. In the meantime, please use the citation given in the <a class="reference external" href="https://github.com/hackingmaterials/matminer">matminer repo</a>.</p>
+</div>
+<div class="section" id="contributing">
+<h1>Contributing<a class="headerlink" href="#contributing" title="Permalink to this headline">¶</a></h1>
+<p>Interested in contributing? See our <a class="reference external" href="https://github.com/hackingmaterials/automatminer/blob/master/CONTRIBUTING.md">contribution guidelines</a> and make a pull request! Please submit questions, issues / bug reports, and all other communication through the  <a class="reference external" href="https://groups.google.com/forum/#!forum/matminer">matminer Google Group</a>.</p>
 </div>
 <div class="section" id="indices-and-tables">
-<h2>Indices and tables<a class="headerlink" href="#indices-and-tables" title="Permalink to this headline">¶</a></h2>
+<h1>Indices and tables<a class="headerlink" href="#indices-and-tables" title="Permalink to this headline">¶</a></h1>
 <ul class="simple">
 <li><a class="reference internal" href="genindex.html"><span class="std std-ref">Index</span></a></li>
 <li><a class="reference internal" href="py-modindex.html"><span class="std std-ref">Module Index</span></a></li>
 <li><a class="reference internal" href="search.html"><span class="std std-ref">Search Page</span></a></li>
 </ul>
-</div>
 </div>