Multioutput classification in Keras - how to get multivariate probabilities and deal with unseen classes












1












$begingroup$


I'm struggling to design in Keras a deep neural network for multioutput classification model. The network works in tandem with external logic in a kind of feedback loop: in each iteration the external module generates the training set, on which the network is trained and then in next iteration the network supports the module in another round of training set generation. I'm rather at rookie+ level and I have encountered 2 problems I do not know how to deal with:



1) how to make a network predict vectors that will not show up in training set, but still account for underlying correlations between vector components



2) how to yield a multivariate distribution on multioutput space.



To be more specific, the "ground truth" of the model is $N$-length vector of binary labels: $y in { 0, 1}^{N} $. This represents some yes/no $N$-dimensional decision and is generated by the external module. The key issue is the yes/no components are highly correlated, $N$ maybe large and the groud truth is not "deterministic" - i.e. it's just some approximation that is expected to get better over the feedback learning process described above.



So far I have designed a simple classification network based on Keras sequence model with $N$ output layer units.



Restating my problems in more specific way:



ad 1) there are $2^{N}$ possible output vectors ("metaclasses"). As $N$ is large, almost surely not all of them will appear in my training set. But I want my network to be flexible enough to be able to come up with unseen metaclasses but still take care of high correlations between single yes/no components.



ad 2) I would like the network to output a few (say three) most probable vectors. In other words, I would like it to yield a (obviously some cut-off) distribution over vectors.









share







New contributor




micsza is a new contributor to this site. Take care in asking for clarification, commenting, and answering.
Check out our Code of Conduct.







$endgroup$

















    1












    $begingroup$


    I'm struggling to design in Keras a deep neural network for multioutput classification model. The network works in tandem with external logic in a kind of feedback loop: in each iteration the external module generates the training set, on which the network is trained and then in next iteration the network supports the module in another round of training set generation. I'm rather at rookie+ level and I have encountered 2 problems I do not know how to deal with:



    1) how to make a network predict vectors that will not show up in training set, but still account for underlying correlations between vector components



    2) how to yield a multivariate distribution on multioutput space.



    To be more specific, the "ground truth" of the model is $N$-length vector of binary labels: $y in { 0, 1}^{N} $. This represents some yes/no $N$-dimensional decision and is generated by the external module. The key issue is the yes/no components are highly correlated, $N$ maybe large and the groud truth is not "deterministic" - i.e. it's just some approximation that is expected to get better over the feedback learning process described above.



    So far I have designed a simple classification network based on Keras sequence model with $N$ output layer units.



    Restating my problems in more specific way:



    ad 1) there are $2^{N}$ possible output vectors ("metaclasses"). As $N$ is large, almost surely not all of them will appear in my training set. But I want my network to be flexible enough to be able to come up with unseen metaclasses but still take care of high correlations between single yes/no components.



    ad 2) I would like the network to output a few (say three) most probable vectors. In other words, I would like it to yield a (obviously some cut-off) distribution over vectors.









    share







    New contributor




    micsza is a new contributor to this site. Take care in asking for clarification, commenting, and answering.
    Check out our Code of Conduct.







    $endgroup$















      1












      1








      1





      $begingroup$


      I'm struggling to design in Keras a deep neural network for multioutput classification model. The network works in tandem with external logic in a kind of feedback loop: in each iteration the external module generates the training set, on which the network is trained and then in next iteration the network supports the module in another round of training set generation. I'm rather at rookie+ level and I have encountered 2 problems I do not know how to deal with:



      1) how to make a network predict vectors that will not show up in training set, but still account for underlying correlations between vector components



      2) how to yield a multivariate distribution on multioutput space.



      To be more specific, the "ground truth" of the model is $N$-length vector of binary labels: $y in { 0, 1}^{N} $. This represents some yes/no $N$-dimensional decision and is generated by the external module. The key issue is the yes/no components are highly correlated, $N$ maybe large and the groud truth is not "deterministic" - i.e. it's just some approximation that is expected to get better over the feedback learning process described above.



      So far I have designed a simple classification network based on Keras sequence model with $N$ output layer units.



      Restating my problems in more specific way:



      ad 1) there are $2^{N}$ possible output vectors ("metaclasses"). As $N$ is large, almost surely not all of them will appear in my training set. But I want my network to be flexible enough to be able to come up with unseen metaclasses but still take care of high correlations between single yes/no components.



      ad 2) I would like the network to output a few (say three) most probable vectors. In other words, I would like it to yield a (obviously some cut-off) distribution over vectors.









      share







      New contributor




      micsza is a new contributor to this site. Take care in asking for clarification, commenting, and answering.
      Check out our Code of Conduct.







      $endgroup$




      I'm struggling to design in Keras a deep neural network for multioutput classification model. The network works in tandem with external logic in a kind of feedback loop: in each iteration the external module generates the training set, on which the network is trained and then in next iteration the network supports the module in another round of training set generation. I'm rather at rookie+ level and I have encountered 2 problems I do not know how to deal with:



      1) how to make a network predict vectors that will not show up in training set, but still account for underlying correlations between vector components



      2) how to yield a multivariate distribution on multioutput space.



      To be more specific, the "ground truth" of the model is $N$-length vector of binary labels: $y in { 0, 1}^{N} $. This represents some yes/no $N$-dimensional decision and is generated by the external module. The key issue is the yes/no components are highly correlated, $N$ maybe large and the groud truth is not "deterministic" - i.e. it's just some approximation that is expected to get better over the feedback learning process described above.



      So far I have designed a simple classification network based on Keras sequence model with $N$ output layer units.



      Restating my problems in more specific way:



      ad 1) there are $2^{N}$ possible output vectors ("metaclasses"). As $N$ is large, almost surely not all of them will appear in my training set. But I want my network to be flexible enough to be able to come up with unseen metaclasses but still take care of high correlations between single yes/no components.



      ad 2) I would like the network to output a few (say three) most probable vectors. In other words, I would like it to yield a (obviously some cut-off) distribution over vectors.







      keras multilabel-classification





      share







      New contributor




      micsza is a new contributor to this site. Take care in asking for clarification, commenting, and answering.
      Check out our Code of Conduct.










      share







      New contributor




      micsza is a new contributor to this site. Take care in asking for clarification, commenting, and answering.
      Check out our Code of Conduct.








      share



      share






      New contributor




      micsza is a new contributor to this site. Take care in asking for clarification, commenting, and answering.
      Check out our Code of Conduct.









      asked 14 hours ago









      micszamicsza

      61




      61




      New contributor




      micsza is a new contributor to this site. Take care in asking for clarification, commenting, and answering.
      Check out our Code of Conduct.





      New contributor





      micsza is a new contributor to this site. Take care in asking for clarification, commenting, and answering.
      Check out our Code of Conduct.






      micsza is a new contributor to this site. Take care in asking for clarification, commenting, and answering.
      Check out our Code of Conduct.






















          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
          });


          }
          });






          micsza is a new contributor. Be nice, and check out our Code of Conduct.










          draft saved

          draft discarded


















          StackExchange.ready(
          function () {
          StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fdatascience.stackexchange.com%2fquestions%2f45312%2fmultioutput-classification-in-keras-how-to-get-multivariate-probabilities-and%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








          micsza is a new contributor. Be nice, and check out our Code of Conduct.










          draft saved

          draft discarded


















          micsza is a new contributor. Be nice, and check out our Code of Conduct.













          micsza is a new contributor. Be nice, and check out our Code of Conduct.












          micsza is a new contributor. Be nice, and check out our Code of Conduct.
















          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%2f45312%2fmultioutput-classification-in-keras-how-to-get-multivariate-probabilities-and%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