<html>
  <head>
    <meta content="text/html; charset=windows-1252"
      http-equiv="Content-Type">
  </head>
  <body bgcolor="#FFFFFF" text="#000000">
    <br>
    <br>
    <div class="moz-cite-prefix">On 03/19/2017 03:47 PM, Thomas
      Evangelidis wrote:<br>
    </div>
    <blockquote
cite="mid:CAACvdx17Ev3jr0ds2bLyJc0RqZkqJH7Rtx=s1ZaodmUvCkcB8Q@mail.gmail.com"
      type="cite">
      <div dir="ltr">
        <div class="gmail_default"
          style="font-size:large;color:rgb(0,0,0)">Which of the
          following methods would you recommend to select good features
          (<=50) from a set of 534 features in order to train a
          MLPregressor? Please take into account that the datasets I use
          for training are small.<br>
        </div>
        <div class="gmail_default"
          style="font-size:large;color:rgb(0,0,0)"><br>
          <a moz-do-not-send="true"
            href="http://scikit-learn.org/stable/modules/feature_selection.html">http://scikit-learn.org/stable/modules/feature_selection.html</a><br>
          <br>
        </div>
        <div class="gmail_default"
          style="font-size:large;color:rgb(0,0,0)">And please don't tell
          me to use a neural network that supports the dropout or any
          other algorithm for feature elimination. This is not
          applicable in my case because I want to know the best 50
          features in order to append them to other types of feature
          that I am confident that are important.<br>
        </div>
        <br>
      </div>
    </blockquote>
    You can always use forward or backward selection as implemented in
    mlxtend if you're patient. As your dataset is small that might work.<br>
    However, it might be hard tricky to get the MLP to run consistently
    - though maybe not...<br>
  </body>
</html>