Reproduce PySM 2 Synchrotron model s1

[zonca]

Implementation of the PySM 2 Synchrotron model, we could easily extend this to a higher N_side

Review and comment in the notebook at https://app.reviewnb.com/healpy/pysm/pull/71/files/

View the notebook at https://nbviewer.jupyter.org/github/healpy/pysm/blob/b42d9e86fcdcf79740f751703ea8e03c0891bb4a/docs/preprocess-templates/reproduce_pysm2_sync_pol.ipynb

[review-notebook-app[bot]]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[review-notebook-app[bot]]

View / edit / reply to this conversation on ReviewNB

zonca commented on 2020-12-15T02:14:30Z ----------------------------------------------------------------

there is a significant difference, I get a steeper EE and a flatter BB

bthorne93 commented on 2020-12-15T19:51:24Z ----------------------------------------------------------------

I'm not sure that the fitting process being done here directly replicates what we did in the paper. We also "fitted" for the white noise amplitude by considering the larger range of multipoles above $\ell_\ast$, and below multipoles of 10. I think if you tried to account for the noise bias, the BB will steepen significantly. As for the EE, I think accounting for sample variance will reduce the significance of the steeper low-ell modes and bring it down.

zonca commented on 2020-12-16T06:59:19Z ----------------------------------------------------------------

I skipped the step where you fit for $\ell_*$ and just used the value from the paper.

would you be able to provide an example on how I should implement this? or provide an improved version of this notebook?

review-notebook-app[bot] commented 3 years ago

View / edit / reply to this conversation on ReviewNB

zonca commented on 2020-12-15T02:14:31Z ----------------------------------------------------------------

the paper mentions as smoothing both 3 degrees and 180/ell_star (5 degrees), not sure where to use the 3 degrees.

I do the fit on the spectra without smoothing because I want to check by eye where the noise picks up.

After it, to make the junction between large and small scales, I directly use 5 degrees, we could probably use 3 degrees here instead of 5. I can compare the 2 versions and see if there is residual noise in the 3 degrees version.

review-notebook-app[bot] commented 3 years ago

View / edit / reply to this conversation on ReviewNB

zonca commented on 2020-12-15T02:14:31Z ----------------------------------------------------------------

patch by patch spectra seem quite well behaved, not sure I need to use spice instead of anafast



review-notebook-app[bot] commented 3 years ago

View / edit / reply to this conversation on ReviewNB

zonca commented on 2020-12-15T02:14:32Z ----------------------------------------------------------------

the difference in the output is mostly due to the different fit to the input maps

b-thorne commented 3 years ago

I'm not sure that the fitting process being done here directly replicates what we did in the paper. We also "fitted" for the white noise amplitude by considering the larger range of multipoles above $\ell_\ast$, and below multipoles of 10. I think if you tried to account for the noise bias, the BB will steepen significantly. As for the EE, I think accounting for sample variance will reduce the significance of the steeper low-ell modes and bring it down.

---

View entire conversation on ReviewNB

zonca commented 3 years ago

I skipped the step where you fit for $\ell_*$ and just used the value from the paper.

would you be able to provide an example on how I should implement this? or provide an improved version of this notebook?

---

View entire conversation on ReviewNB

zonca commented 3 years ago

ok, I implemented suggestions by @bthorne93, they worked fine for dust, for synchrotron there are still some differences. I don't think it is useful for me to go deeper than this. I am going to merge this in the docs, the notebook tries to show and explain the differences. @NicolettaK @brandonshensley in case anyone is interested in pursuing this further, there will also be a link to the full notebook.

• © Githubissues.
• Githubissues is a development platform for aggregating issues.

[review-notebook-app[bot]]

View / edit / reply to this conversation on ReviewNB

zonca commented on 2020-12-15T02:14:31Z ----------------------------------------------------------------

the paper mentions as smoothing both 3 degrees and 180/ell_star (5 degrees), not sure where to use the 3 degrees.

I do the fit on the spectra without smoothing because I want to check by eye where the noise picks up.

After it, to make the junction between large and small scales, I directly use 5 degrees, we could probably use 3 degrees here instead of 5. I can compare the 2 versions and see if there is residual noise in the 3 degrees version.

review-notebook-app[bot] commented 3 years ago

View / edit / reply to this conversation on ReviewNB

zonca commented on 2020-12-15T02:14:31Z ----------------------------------------------------------------

patch by patch spectra seem quite well behaved, not sure I need to use spice instead of anafast



review-notebook-app[bot] commented 3 years ago

View / edit / reply to this conversation on ReviewNB

zonca commented on 2020-12-15T02:14:32Z ----------------------------------------------------------------

the difference in the output is mostly due to the different fit to the input maps

b-thorne commented 3 years ago

I'm not sure that the fitting process being done here directly replicates what we did in the paper. We also "fitted" for the white noise amplitude by considering the larger range of multipoles above $\ell_\ast$, and below multipoles of 10. I think if you tried to account for the noise bias, the BB will steepen significantly. As for the EE, I think accounting for sample variance will reduce the significance of the steeper low-ell modes and bring it down.

---

View entire conversation on ReviewNB

zonca commented 3 years ago

I skipped the step where you fit for $\ell_*$ and just used the value from the paper.

would you be able to provide an example on how I should implement this? or provide an improved version of this notebook?

---

View entire conversation on ReviewNB

zonca commented 3 years ago

ok, I implemented suggestions by @bthorne93, they worked fine for dust, for synchrotron there are still some differences. I don't think it is useful for me to go deeper than this. I am going to merge this in the docs, the notebook tries to show and explain the differences. @NicolettaK @brandonshensley in case anyone is interested in pursuing this further, there will also be a link to the full notebook.

• © Githubissues.
• Githubissues is a development platform for aggregating issues.

[review-notebook-app[bot]]

View / edit / reply to this conversation on ReviewNB

zonca commented on 2020-12-15T02:14:31Z ----------------------------------------------------------------

patch by patch spectra seem quite well behaved, not sure I need to use spice instead of anafast



review-notebook-app[bot] commented 3 years ago

View / edit / reply to this conversation on ReviewNB

zonca commented on 2020-12-15T02:14:32Z ----------------------------------------------------------------

the difference in the output is mostly due to the different fit to the input maps

b-thorne commented 3 years ago

I'm not sure that the fitting process being done here directly replicates what we did in the paper. We also "fitted" for the white noise amplitude by considering the larger range of multipoles above $\ell_\ast$, and below multipoles of 10. I think if you tried to account for the noise bias, the BB will steepen significantly. As for the EE, I think accounting for sample variance will reduce the significance of the steeper low-ell modes and bring it down.

---

View entire conversation on ReviewNB

zonca commented 3 years ago

I skipped the step where you fit for $\ell_*$ and just used the value from the paper.

would you be able to provide an example on how I should implement this? or provide an improved version of this notebook?

---

View entire conversation on ReviewNB

zonca commented 3 years ago

ok, I implemented suggestions by @bthorne93, they worked fine for dust, for synchrotron there are still some differences. I don't think it is useful for me to go deeper than this. I am going to merge this in the docs, the notebook tries to show and explain the differences. @NicolettaK @brandonshensley in case anyone is interested in pursuing this further, there will also be a link to the full notebook.

• © Githubissues.
• Githubissues is a development platform for aggregating issues.

[review-notebook-app[bot]]

View / edit / reply to this conversation on ReviewNB

zonca commented on 2020-12-15T02:14:32Z ----------------------------------------------------------------

the difference in the output is mostly due to the different fit to the input maps

b-thorne commented 3 years ago

I'm not sure that the fitting process being done here directly replicates what we did in the paper. We also "fitted" for the white noise amplitude by considering the larger range of multipoles above $\ell_\ast$, and below multipoles of 10. I think if you tried to account for the noise bias, the BB will steepen significantly. As for the EE, I think accounting for sample variance will reduce the significance of the steeper low-ell modes and bring it down.

---

View entire conversation on ReviewNB

zonca commented 3 years ago

I skipped the step where you fit for $\ell_*$ and just used the value from the paper.

would you be able to provide an example on how I should implement this? or provide an improved version of this notebook?

---

View entire conversation on ReviewNB

zonca commented 3 years ago

ok, I implemented suggestions by @bthorne93, they worked fine for dust, for synchrotron there are still some differences. I don't think it is useful for me to go deeper than this. I am going to merge this in the docs, the notebook tries to show and explain the differences. @NicolettaK @brandonshensley in case anyone is interested in pursuing this further, there will also be a link to the full notebook.

• © Githubissues.
• Githubissues is a development platform for aggregating issues.

[b-thorne]

I'm not sure that the fitting process being done here directly replicates what we did in the paper. We also "fitted" for the white noise amplitude by considering the larger range of multipoles above $\ell_\ast$, and below multipoles of 10. I think if you tried to account for the noise bias, the BB will steepen significantly. As for the EE, I think accounting for sample variance will reduce the significance of the steeper low-ell modes and bring it down.

---

View entire conversation on ReviewNB

[zonca]

I skipped the step where you fit for $\ell_*$ and just used the value from the paper.

would you be able to provide an example on how I should implement this? or provide an improved version of this notebook?

---

View entire conversation on ReviewNB

[zonca]

ok, I implemented suggestions by @bthorne93, they worked fine for dust, for synchrotron there are still some differences. I don't think it is useful for me to go deeper than this. I am going to merge this in the docs, the notebook tries to show and explain the differences. @NicolettaK @brandonshensley in case anyone is interested in pursuing this further, there will also be a link to the full notebook.