Continuous Tikhonov Regularization for Deconvolution











up vote
4
down vote

favorite












I am trying to solve the following deconvolution problem where $g(s)$ is a known real valued function and has finite energy:



$$g(s) = int_{-infty}^{infty} frac{1}{sqrt{2pi}}f(t)e^{-(t-s)^2/2}dt$$



Since the integral kernel is a difference kernel the solution is available via convolution theorem.



$$f(t) = F^{-1}left[frac{F[g](w)}{e^{-w^2/2}}right]$$



where $F$ denotes the fourier transform.



Now on top of this I would like to impose finite energy condition on $f$, so I wrote the following regularized error minimization



$$hat{f}=text{argmin}_f left(int_{-infty}^{infty}[g(s) - int_{-infty}^{infty} frac{1}{sqrt{2pi}}f(t)e^{-(t-s)^2/2}dt]^2 ds+ alphaint_{-infty}^{infty} |f(t)|^2dtright)$$



I would like to show that:



$$hat{f}(t) = F^{-1}left[frac{F[g](w)}{e^{-w^2/2}+alpha}right]$$



or something in the line of this.










share|cite|improve this question




























    up vote
    4
    down vote

    favorite












    I am trying to solve the following deconvolution problem where $g(s)$ is a known real valued function and has finite energy:



    $$g(s) = int_{-infty}^{infty} frac{1}{sqrt{2pi}}f(t)e^{-(t-s)^2/2}dt$$



    Since the integral kernel is a difference kernel the solution is available via convolution theorem.



    $$f(t) = F^{-1}left[frac{F[g](w)}{e^{-w^2/2}}right]$$



    where $F$ denotes the fourier transform.



    Now on top of this I would like to impose finite energy condition on $f$, so I wrote the following regularized error minimization



    $$hat{f}=text{argmin}_f left(int_{-infty}^{infty}[g(s) - int_{-infty}^{infty} frac{1}{sqrt{2pi}}f(t)e^{-(t-s)^2/2}dt]^2 ds+ alphaint_{-infty}^{infty} |f(t)|^2dtright)$$



    I would like to show that:



    $$hat{f}(t) = F^{-1}left[frac{F[g](w)}{e^{-w^2/2}+alpha}right]$$



    or something in the line of this.










    share|cite|improve this question


























      up vote
      4
      down vote

      favorite









      up vote
      4
      down vote

      favorite











      I am trying to solve the following deconvolution problem where $g(s)$ is a known real valued function and has finite energy:



      $$g(s) = int_{-infty}^{infty} frac{1}{sqrt{2pi}}f(t)e^{-(t-s)^2/2}dt$$



      Since the integral kernel is a difference kernel the solution is available via convolution theorem.



      $$f(t) = F^{-1}left[frac{F[g](w)}{e^{-w^2/2}}right]$$



      where $F$ denotes the fourier transform.



      Now on top of this I would like to impose finite energy condition on $f$, so I wrote the following regularized error minimization



      $$hat{f}=text{argmin}_f left(int_{-infty}^{infty}[g(s) - int_{-infty}^{infty} frac{1}{sqrt{2pi}}f(t)e^{-(t-s)^2/2}dt]^2 ds+ alphaint_{-infty}^{infty} |f(t)|^2dtright)$$



      I would like to show that:



      $$hat{f}(t) = F^{-1}left[frac{F[g](w)}{e^{-w^2/2}+alpha}right]$$



      or something in the line of this.










      share|cite|improve this question















      I am trying to solve the following deconvolution problem where $g(s)$ is a known real valued function and has finite energy:



      $$g(s) = int_{-infty}^{infty} frac{1}{sqrt{2pi}}f(t)e^{-(t-s)^2/2}dt$$



      Since the integral kernel is a difference kernel the solution is available via convolution theorem.



      $$f(t) = F^{-1}left[frac{F[g](w)}{e^{-w^2/2}}right]$$



      where $F$ denotes the fourier transform.



      Now on top of this I would like to impose finite energy condition on $f$, so I wrote the following regularized error minimization



      $$hat{f}=text{argmin}_f left(int_{-infty}^{infty}[g(s) - int_{-infty}^{infty} frac{1}{sqrt{2pi}}f(t)e^{-(t-s)^2/2}dt]^2 ds+ alphaint_{-infty}^{infty} |f(t)|^2dtright)$$



      I would like to show that:



      $$hat{f}(t) = F^{-1}left[frac{F[g](w)}{e^{-w^2/2}+alpha}right]$$



      or something in the line of this.







      fourier-transform convolution regularization






      share|cite|improve this question















      share|cite|improve this question













      share|cite|improve this question




      share|cite|improve this question








      edited Nov 15 at 13:09









      mathreadler

      14.6k72160




      14.6k72160










      asked Nov 15 at 10:07









      Cowboy Trader

      8712




      8712






















          1 Answer
          1






          active

          oldest

          votes

















          up vote
          2
          down vote













          Yes, this is broadly correct and the main step that you need is that the Fourier transformation is an isometry so $int_{-infty}^infty |f(t)|^2 , dt = int_{-infty}^infty |F(omega)|^2 , domega $ (up to a factor of $2pi$). So, Fourier transform the problem and you can write the objective in Fourier space. The result then follows straightaway.






          share|cite|improve this answer





















          • Just doing a Fourier transformation right away without regularization is not very advisable in presence of noise.
            – mathreadler
            Nov 15 at 13:09










          • That's not what I'm saying. Write the objective function (of the regularised problem) in Fourier space, and solve for $F(omega)$. In fact, a glance at the answer shows you that $F(omega)=G(omega)/[e^{-omega^2/2} + a]$.
            – Richard Martin
            Nov 15 at 13:21












          • I think the answer is probably going to look like $F[K]G(w)/(F[K]F[K] + alpha)$ where F[K] is the fourier of kernel, which resembles more the typical tikhonov regularized least squares formula.
            – Cowboy Trader
            Nov 15 at 13:23












          • @CowboyTrader No this is not good way to do it. You need more freedom for regularization and adding more information to avoid underdetermination.
            – mathreadler
            Nov 15 at 13:25










          • Effectively all you're doing is this. You don't want to divide by $e^{-omega^2/2}$ because it comes too close to zero (when $|omega|$ large). But if you add a constant on to the denominator, then it becomes bounded away from zero. Job done. The same with any situation where you need to "divide" by a matrix $M$ whose eigenvalues might get close to zero: replace the matrix with $M+epsilon I$.
            – Richard Martin
            Nov 15 at 13:46













          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: "69"
          };
          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',
          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
          },
          noCode: true, onDemand: true,
          discardSelector: ".discard-answer"
          ,immediatelyShowMarkdownHelp:true
          });


          }
          });














          draft saved

          draft discarded


















          StackExchange.ready(
          function () {
          StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fmath.stackexchange.com%2fquestions%2f2999487%2fcontinuous-tikhonov-regularization-for-deconvolution%23new-answer', 'question_page');
          }
          );

          Post as a guest















          Required, but never shown

























          1 Answer
          1






          active

          oldest

          votes








          1 Answer
          1






          active

          oldest

          votes









          active

          oldest

          votes






          active

          oldest

          votes








          up vote
          2
          down vote













          Yes, this is broadly correct and the main step that you need is that the Fourier transformation is an isometry so $int_{-infty}^infty |f(t)|^2 , dt = int_{-infty}^infty |F(omega)|^2 , domega $ (up to a factor of $2pi$). So, Fourier transform the problem and you can write the objective in Fourier space. The result then follows straightaway.






          share|cite|improve this answer





















          • Just doing a Fourier transformation right away without regularization is not very advisable in presence of noise.
            – mathreadler
            Nov 15 at 13:09










          • That's not what I'm saying. Write the objective function (of the regularised problem) in Fourier space, and solve for $F(omega)$. In fact, a glance at the answer shows you that $F(omega)=G(omega)/[e^{-omega^2/2} + a]$.
            – Richard Martin
            Nov 15 at 13:21












          • I think the answer is probably going to look like $F[K]G(w)/(F[K]F[K] + alpha)$ where F[K] is the fourier of kernel, which resembles more the typical tikhonov regularized least squares formula.
            – Cowboy Trader
            Nov 15 at 13:23












          • @CowboyTrader No this is not good way to do it. You need more freedom for regularization and adding more information to avoid underdetermination.
            – mathreadler
            Nov 15 at 13:25










          • Effectively all you're doing is this. You don't want to divide by $e^{-omega^2/2}$ because it comes too close to zero (when $|omega|$ large). But if you add a constant on to the denominator, then it becomes bounded away from zero. Job done. The same with any situation where you need to "divide" by a matrix $M$ whose eigenvalues might get close to zero: replace the matrix with $M+epsilon I$.
            – Richard Martin
            Nov 15 at 13:46

















          up vote
          2
          down vote













          Yes, this is broadly correct and the main step that you need is that the Fourier transformation is an isometry so $int_{-infty}^infty |f(t)|^2 , dt = int_{-infty}^infty |F(omega)|^2 , domega $ (up to a factor of $2pi$). So, Fourier transform the problem and you can write the objective in Fourier space. The result then follows straightaway.






          share|cite|improve this answer





















          • Just doing a Fourier transformation right away without regularization is not very advisable in presence of noise.
            – mathreadler
            Nov 15 at 13:09










          • That's not what I'm saying. Write the objective function (of the regularised problem) in Fourier space, and solve for $F(omega)$. In fact, a glance at the answer shows you that $F(omega)=G(omega)/[e^{-omega^2/2} + a]$.
            – Richard Martin
            Nov 15 at 13:21












          • I think the answer is probably going to look like $F[K]G(w)/(F[K]F[K] + alpha)$ where F[K] is the fourier of kernel, which resembles more the typical tikhonov regularized least squares formula.
            – Cowboy Trader
            Nov 15 at 13:23












          • @CowboyTrader No this is not good way to do it. You need more freedom for regularization and adding more information to avoid underdetermination.
            – mathreadler
            Nov 15 at 13:25










          • Effectively all you're doing is this. You don't want to divide by $e^{-omega^2/2}$ because it comes too close to zero (when $|omega|$ large). But if you add a constant on to the denominator, then it becomes bounded away from zero. Job done. The same with any situation where you need to "divide" by a matrix $M$ whose eigenvalues might get close to zero: replace the matrix with $M+epsilon I$.
            – Richard Martin
            Nov 15 at 13:46















          up vote
          2
          down vote










          up vote
          2
          down vote









          Yes, this is broadly correct and the main step that you need is that the Fourier transformation is an isometry so $int_{-infty}^infty |f(t)|^2 , dt = int_{-infty}^infty |F(omega)|^2 , domega $ (up to a factor of $2pi$). So, Fourier transform the problem and you can write the objective in Fourier space. The result then follows straightaway.






          share|cite|improve this answer












          Yes, this is broadly correct and the main step that you need is that the Fourier transformation is an isometry so $int_{-infty}^infty |f(t)|^2 , dt = int_{-infty}^infty |F(omega)|^2 , domega $ (up to a factor of $2pi$). So, Fourier transform the problem and you can write the objective in Fourier space. The result then follows straightaway.







          share|cite|improve this answer












          share|cite|improve this answer



          share|cite|improve this answer










          answered Nov 15 at 10:29









          Richard Martin

          1,6148




          1,6148












          • Just doing a Fourier transformation right away without regularization is not very advisable in presence of noise.
            – mathreadler
            Nov 15 at 13:09










          • That's not what I'm saying. Write the objective function (of the regularised problem) in Fourier space, and solve for $F(omega)$. In fact, a glance at the answer shows you that $F(omega)=G(omega)/[e^{-omega^2/2} + a]$.
            – Richard Martin
            Nov 15 at 13:21












          • I think the answer is probably going to look like $F[K]G(w)/(F[K]F[K] + alpha)$ where F[K] is the fourier of kernel, which resembles more the typical tikhonov regularized least squares formula.
            – Cowboy Trader
            Nov 15 at 13:23












          • @CowboyTrader No this is not good way to do it. You need more freedom for regularization and adding more information to avoid underdetermination.
            – mathreadler
            Nov 15 at 13:25










          • Effectively all you're doing is this. You don't want to divide by $e^{-omega^2/2}$ because it comes too close to zero (when $|omega|$ large). But if you add a constant on to the denominator, then it becomes bounded away from zero. Job done. The same with any situation where you need to "divide" by a matrix $M$ whose eigenvalues might get close to zero: replace the matrix with $M+epsilon I$.
            – Richard Martin
            Nov 15 at 13:46




















          • Just doing a Fourier transformation right away without regularization is not very advisable in presence of noise.
            – mathreadler
            Nov 15 at 13:09










          • That's not what I'm saying. Write the objective function (of the regularised problem) in Fourier space, and solve for $F(omega)$. In fact, a glance at the answer shows you that $F(omega)=G(omega)/[e^{-omega^2/2} + a]$.
            – Richard Martin
            Nov 15 at 13:21












          • I think the answer is probably going to look like $F[K]G(w)/(F[K]F[K] + alpha)$ where F[K] is the fourier of kernel, which resembles more the typical tikhonov regularized least squares formula.
            – Cowboy Trader
            Nov 15 at 13:23












          • @CowboyTrader No this is not good way to do it. You need more freedom for regularization and adding more information to avoid underdetermination.
            – mathreadler
            Nov 15 at 13:25










          • Effectively all you're doing is this. You don't want to divide by $e^{-omega^2/2}$ because it comes too close to zero (when $|omega|$ large). But if you add a constant on to the denominator, then it becomes bounded away from zero. Job done. The same with any situation where you need to "divide" by a matrix $M$ whose eigenvalues might get close to zero: replace the matrix with $M+epsilon I$.
            – Richard Martin
            Nov 15 at 13:46


















          Just doing a Fourier transformation right away without regularization is not very advisable in presence of noise.
          – mathreadler
          Nov 15 at 13:09




          Just doing a Fourier transformation right away without regularization is not very advisable in presence of noise.
          – mathreadler
          Nov 15 at 13:09












          That's not what I'm saying. Write the objective function (of the regularised problem) in Fourier space, and solve for $F(omega)$. In fact, a glance at the answer shows you that $F(omega)=G(omega)/[e^{-omega^2/2} + a]$.
          – Richard Martin
          Nov 15 at 13:21






          That's not what I'm saying. Write the objective function (of the regularised problem) in Fourier space, and solve for $F(omega)$. In fact, a glance at the answer shows you that $F(omega)=G(omega)/[e^{-omega^2/2} + a]$.
          – Richard Martin
          Nov 15 at 13:21














          I think the answer is probably going to look like $F[K]G(w)/(F[K]F[K] + alpha)$ where F[K] is the fourier of kernel, which resembles more the typical tikhonov regularized least squares formula.
          – Cowboy Trader
          Nov 15 at 13:23






          I think the answer is probably going to look like $F[K]G(w)/(F[K]F[K] + alpha)$ where F[K] is the fourier of kernel, which resembles more the typical tikhonov regularized least squares formula.
          – Cowboy Trader
          Nov 15 at 13:23














          @CowboyTrader No this is not good way to do it. You need more freedom for regularization and adding more information to avoid underdetermination.
          – mathreadler
          Nov 15 at 13:25




          @CowboyTrader No this is not good way to do it. You need more freedom for regularization and adding more information to avoid underdetermination.
          – mathreadler
          Nov 15 at 13:25












          Effectively all you're doing is this. You don't want to divide by $e^{-omega^2/2}$ because it comes too close to zero (when $|omega|$ large). But if you add a constant on to the denominator, then it becomes bounded away from zero. Job done. The same with any situation where you need to "divide" by a matrix $M$ whose eigenvalues might get close to zero: replace the matrix with $M+epsilon I$.
          – Richard Martin
          Nov 15 at 13:46






          Effectively all you're doing is this. You don't want to divide by $e^{-omega^2/2}$ because it comes too close to zero (when $|omega|$ large). But if you add a constant on to the denominator, then it becomes bounded away from zero. Job done. The same with any situation where you need to "divide" by a matrix $M$ whose eigenvalues might get close to zero: replace the matrix with $M+epsilon I$.
          – Richard Martin
          Nov 15 at 13:46




















          draft saved

          draft discarded




















































          Thanks for contributing an answer to Mathematics 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.





          Some of your past answers have not been well-received, and you're in danger of being blocked from answering.


          Please pay close attention to the following guidance:


          • 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%2fmath.stackexchange.com%2fquestions%2f2999487%2fcontinuous-tikhonov-regularization-for-deconvolution%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

          Biblatex bibliography style without URLs when DOI exists (in Overleaf with Zotero bibliography)

          ComboBox Display Member on multiple fields

          Is it possible to collect Nectar points via Trainline?