Ways to normalize multiple timeseries samples for shallow/deep learning?












0












$begingroup$


Let's say I have many samples that are timeseries, giving me data with shape (samples, timesteps, features). What would be the correct way of normalizing them, if the raw values can vary by orders of magnitude due to e.g. different instruments which output data in various scales (such as recording with different bit depths).



For deep learning, I guess it would make sense just to normalize each individual sample to [0,1] (or [-1,1] if timeseries is supposed to be stationary?), given that that the sample has a natural min/max, and that that with batch normalization and activations, the network doesn't care.



But for standard classifiers, if datapoints are shuffled and time dependence removed, which way makes the most sense? Reshaping to (samples, features) and standardizing? Or normalizing each individual sample, then reshaping, thus removing the need for a scaler altogether?



I guess having data raw data in [0, 2^8] and [0, 2^16], then normalizing them together would wreck different amounts of havoc on various algorithms, because they violate some underlying assumption about uniform/normal distribution of features?










share|improve this question









$endgroup$

















    0












    $begingroup$


    Let's say I have many samples that are timeseries, giving me data with shape (samples, timesteps, features). What would be the correct way of normalizing them, if the raw values can vary by orders of magnitude due to e.g. different instruments which output data in various scales (such as recording with different bit depths).



    For deep learning, I guess it would make sense just to normalize each individual sample to [0,1] (or [-1,1] if timeseries is supposed to be stationary?), given that that the sample has a natural min/max, and that that with batch normalization and activations, the network doesn't care.



    But for standard classifiers, if datapoints are shuffled and time dependence removed, which way makes the most sense? Reshaping to (samples, features) and standardizing? Or normalizing each individual sample, then reshaping, thus removing the need for a scaler altogether?



    I guess having data raw data in [0, 2^8] and [0, 2^16], then normalizing them together would wreck different amounts of havoc on various algorithms, because they violate some underlying assumption about uniform/normal distribution of features?










    share|improve this question









    $endgroup$















      0












      0








      0





      $begingroup$


      Let's say I have many samples that are timeseries, giving me data with shape (samples, timesteps, features). What would be the correct way of normalizing them, if the raw values can vary by orders of magnitude due to e.g. different instruments which output data in various scales (such as recording with different bit depths).



      For deep learning, I guess it would make sense just to normalize each individual sample to [0,1] (or [-1,1] if timeseries is supposed to be stationary?), given that that the sample has a natural min/max, and that that with batch normalization and activations, the network doesn't care.



      But for standard classifiers, if datapoints are shuffled and time dependence removed, which way makes the most sense? Reshaping to (samples, features) and standardizing? Or normalizing each individual sample, then reshaping, thus removing the need for a scaler altogether?



      I guess having data raw data in [0, 2^8] and [0, 2^16], then normalizing them together would wreck different amounts of havoc on various algorithms, because they violate some underlying assumption about uniform/normal distribution of features?










      share|improve this question









      $endgroup$




      Let's say I have many samples that are timeseries, giving me data with shape (samples, timesteps, features). What would be the correct way of normalizing them, if the raw values can vary by orders of magnitude due to e.g. different instruments which output data in various scales (such as recording with different bit depths).



      For deep learning, I guess it would make sense just to normalize each individual sample to [0,1] (or [-1,1] if timeseries is supposed to be stationary?), given that that the sample has a natural min/max, and that that with batch normalization and activations, the network doesn't care.



      But for standard classifiers, if datapoints are shuffled and time dependence removed, which way makes the most sense? Reshaping to (samples, features) and standardizing? Or normalizing each individual sample, then reshaping, thus removing the need for a scaler altogether?



      I guess having data raw data in [0, 2^8] and [0, 2^16], then normalizing them together would wreck different amounts of havoc on various algorithms, because they violate some underlying assumption about uniform/normal distribution of features?







      time-series normalization






      share|improve this question













      share|improve this question











      share|improve this question




      share|improve this question










      asked 2 days ago









      komodovaran_komodovaran_

      1113




      1113






















          0






          active

          oldest

          votes











          Your Answer





          StackExchange.ifUsing("editor", function () {
          return StackExchange.using("mathjaxEditing", function () {
          StackExchange.MarkdownEditor.creationCallbacks.add(function (editor, postfix) {
          StackExchange.mathjaxEditing.prepareWmdForMathJax(editor, postfix, [["$", "$"], ["\\(","\\)"]]);
          });
          });
          }, "mathjax-editing");

          StackExchange.ready(function() {
          var channelOptions = {
          tags: "".split(" "),
          id: "557"
          };
          initTagRenderer("".split(" "), "".split(" "), channelOptions);

          StackExchange.using("externalEditor", function() {
          // Have to fire editor after snippets, if snippets enabled
          if (StackExchange.settings.snippets.snippetsEnabled) {
          StackExchange.using("snippets", function() {
          createEditor();
          });
          }
          else {
          createEditor();
          }
          });

          function createEditor() {
          StackExchange.prepareEditor({
          heartbeatType: 'answer',
          autoActivateHeartbeat: false,
          convertImagesToLinks: false,
          noModals: true,
          showLowRepImageUploadWarning: true,
          reputationToPostImages: null,
          bindNavPrevention: true,
          postfix: "",
          imageUploader: {
          brandingHtml: "Powered by u003ca class="icon-imgur-white" href="https://imgur.com/"u003eu003c/au003e",
          contentPolicyHtml: "User contributions licensed under u003ca href="https://creativecommons.org/licenses/by-sa/3.0/"u003ecc by-sa 3.0 with attribution requiredu003c/au003e u003ca href="https://stackoverflow.com/legal/content-policy"u003e(content policy)u003c/au003e",
          allowUrls: true
          },
          onDemand: true,
          discardSelector: ".discard-answer"
          ,immediatelyShowMarkdownHelp:true
          });


          }
          });














          draft saved

          draft discarded


















          StackExchange.ready(
          function () {
          StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fdatascience.stackexchange.com%2fquestions%2f46639%2fways-to-normalize-multiple-timeseries-samples-for-shallow-deep-learning%23new-answer', 'question_page');
          }
          );

          Post as a guest















          Required, but never shown

























          0






          active

          oldest

          votes








          0






          active

          oldest

          votes









          active

          oldest

          votes






          active

          oldest

          votes
















          draft saved

          draft discarded




















































          Thanks for contributing an answer to Data Science Stack Exchange!


          • Please be sure to answer the question. Provide details and share your research!

          But avoid



          • Asking for help, clarification, or responding to other answers.

          • Making statements based on opinion; back them up with references or personal experience.


          Use MathJax to format equations. MathJax reference.


          To learn more, see our tips on writing great answers.




          draft saved


          draft discarded














          StackExchange.ready(
          function () {
          StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fdatascience.stackexchange.com%2fquestions%2f46639%2fways-to-normalize-multiple-timeseries-samples-for-shallow-deep-learning%23new-answer', 'question_page');
          }
          );

          Post as a guest















          Required, but never shown





















































          Required, but never shown














          Required, but never shown












          Required, but never shown







          Required, but never shown

































          Required, but never shown














          Required, but never shown












          Required, but never shown







          Required, but never shown







          Popular posts from this blog

          How to label and detect the document text images

          Vallis Paradisi

          Tabula Rosettana