Scikit-learn pipelines : how to deal with hyperparameters in a clean way? (pipe and filter pattern with...

Multi tool use
Multi tool use












-1















Let's suppose that I have a Pipeline in scikit-learn (pipe and filter pattern).



Now let's suppose that this Pipeline is wrapped in a class of its own and that it has a pre-made list of hyperparameters, or a premade list of hyperparameters grid, as a static class constant. The class stays a pipeline so that it has the .fit and .transform (it inherits from the Pipeline object, but at construction, it already knows which pipeline steps to compose).



I now want pipelines of pipelines, so one master pipeline would compose each one. In the end, I want to perform a hyperparameter search, so I'd need the first pipeline to get the hyperparameters of every sub-pipeline and add the string prefix as needed.



So the question is: How to do this (optional: in a clean way)? I feel that I break some Object Oriented Programming (OOP)'s SOLID principles by having to get one massive hyperparameter grid out of each pipeline somehow recursively and to prepend prefixes to params with some __ separators.



Any suggestion will be appreciated. It's obvious that in my thought experiment, something is wrong with the way to share hyperparameters, and I'd like to fix that with. Can you suggest a design pattern or a more usable way to do that?










share|improve this question



























    -1















    Let's suppose that I have a Pipeline in scikit-learn (pipe and filter pattern).



    Now let's suppose that this Pipeline is wrapped in a class of its own and that it has a pre-made list of hyperparameters, or a premade list of hyperparameters grid, as a static class constant. The class stays a pipeline so that it has the .fit and .transform (it inherits from the Pipeline object, but at construction, it already knows which pipeline steps to compose).



    I now want pipelines of pipelines, so one master pipeline would compose each one. In the end, I want to perform a hyperparameter search, so I'd need the first pipeline to get the hyperparameters of every sub-pipeline and add the string prefix as needed.



    So the question is: How to do this (optional: in a clean way)? I feel that I break some Object Oriented Programming (OOP)'s SOLID principles by having to get one massive hyperparameter grid out of each pipeline somehow recursively and to prepend prefixes to params with some __ separators.



    Any suggestion will be appreciated. It's obvious that in my thought experiment, something is wrong with the way to share hyperparameters, and I'd like to fix that with. Can you suggest a design pattern or a more usable way to do that?










    share|improve this question

























      -1












      -1








      -1








      Let's suppose that I have a Pipeline in scikit-learn (pipe and filter pattern).



      Now let's suppose that this Pipeline is wrapped in a class of its own and that it has a pre-made list of hyperparameters, or a premade list of hyperparameters grid, as a static class constant. The class stays a pipeline so that it has the .fit and .transform (it inherits from the Pipeline object, but at construction, it already knows which pipeline steps to compose).



      I now want pipelines of pipelines, so one master pipeline would compose each one. In the end, I want to perform a hyperparameter search, so I'd need the first pipeline to get the hyperparameters of every sub-pipeline and add the string prefix as needed.



      So the question is: How to do this (optional: in a clean way)? I feel that I break some Object Oriented Programming (OOP)'s SOLID principles by having to get one massive hyperparameter grid out of each pipeline somehow recursively and to prepend prefixes to params with some __ separators.



      Any suggestion will be appreciated. It's obvious that in my thought experiment, something is wrong with the way to share hyperparameters, and I'd like to fix that with. Can you suggest a design pattern or a more usable way to do that?










      share|improve this question














      Let's suppose that I have a Pipeline in scikit-learn (pipe and filter pattern).



      Now let's suppose that this Pipeline is wrapped in a class of its own and that it has a pre-made list of hyperparameters, or a premade list of hyperparameters grid, as a static class constant. The class stays a pipeline so that it has the .fit and .transform (it inherits from the Pipeline object, but at construction, it already knows which pipeline steps to compose).



      I now want pipelines of pipelines, so one master pipeline would compose each one. In the end, I want to perform a hyperparameter search, so I'd need the first pipeline to get the hyperparameters of every sub-pipeline and add the string prefix as needed.



      So the question is: How to do this (optional: in a clean way)? I feel that I break some Object Oriented Programming (OOP)'s SOLID principles by having to get one massive hyperparameter grid out of each pipeline somehow recursively and to prepend prefixes to params with some __ separators.



      Any suggestion will be appreciated. It's obvious that in my thought experiment, something is wrong with the way to share hyperparameters, and I'd like to fix that with. Can you suggest a design pattern or a more usable way to do that?







      oop machine-learning design-patterns scikit-learn pipeline






      share|improve this question













      share|improve this question











      share|improve this question




      share|improve this question










      asked Dec 29 '18 at 21:00









      Guillaume ChevalierGuillaume Chevalier

      2,76512342




      2,76512342
























          0






          active

          oldest

          votes











          Your Answer






          StackExchange.ifUsing("editor", function () {
          StackExchange.using("externalEditor", function () {
          StackExchange.using("snippets", function () {
          StackExchange.snippets.init();
          });
          });
          }, "code-snippets");

          StackExchange.ready(function() {
          var channelOptions = {
          tags: "".split(" "),
          id: "1"
          };
          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: true,
          noModals: true,
          showLowRepImageUploadWarning: true,
          reputationToPostImages: 10,
          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%2fstackoverflow.com%2fquestions%2f53973303%2fscikit-learn-pipelines-how-to-deal-with-hyperparameters-in-a-clean-way-pipe%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 Stack Overflow!


          • 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.


          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%2fstackoverflow.com%2fquestions%2f53973303%2fscikit-learn-pipelines-how-to-deal-with-hyperparameters-in-a-clean-way-pipe%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







          RjtMiOiKWdnD,1kVu
          2 rrU9xSH1dqYhLETVZ

          Popular posts from this blog

          Monofisismo

          Angular Downloading a file using contenturl with Basic Authentication

          Olmecas