[REVIEW]: InvertibleNetworks.jl: A Julia package for scalable normalizing flows

[editorialbot]

Submitting author: !--author-handle-->@rafaelorozco<!--end-author-handle-- (Rafael Orozco) Repository: https://github.com/slimgroup/InvertibleNetworks.jl Branch with paper.md (empty if default branch): paper-joss Version: 2.2.9 Editor: !--editor-->@drvinceknight<!--end-editor-- Reviewers: @aurorarossi, @Nando-Hegemann Archive: 10.5281/zenodo.12810006

Status

Status badge code:

HTML: <a href="https://joss.theoj.org/papers/b1e8bc48e52694bcc9e4fb87634fc39f"><img src="https://joss.theoj.org/papers/b1e8bc48e52694bcc9e4fb87634fc39f/status.svg"></a>
Markdown: [![status](https://joss.theoj.org/papers/b1e8bc48e52694bcc9e4fb87634fc39f/status.svg)](https://joss.theoj.org/papers/b1e8bc48e52694bcc9e4fb87634fc39f)

Reviewers and authors:

Please avoid lengthy details of difficulties in the review thread. Instead, please create a new issue in the target repository and link to those issues (especially acceptance-blockers) by leaving comments in the review thread below. (For completists: if the target issue tracker is also on GitHub, linking the review thread in the issue or vice versa will create corresponding breadcrumb trails in the link target.)

Reviewer instructions & questions

@aurorarossi & @Nando-Hegemann, your review will be checklist based. Each of you will have a separate checklist that you should update when carrying out your review. First of all you need to run this command in a separate comment to create the checklist:

@editorialbot generate my checklist

The reviewer guidelines are available here: https://joss.readthedocs.io/en/latest/reviewer_guidelines.html. Any questions/concerns please let @drvinceknight know.

✨ Please start on your review when you are able, and be sure to complete your review in the next six weeks, at the very latest ✨

Checklists

📝 Checklist for @aurorarossi

📝 Checklist for @Nando-Hegemann

[editorialbot]

Hello humans, I'm @editorialbot, a robot that can help you with some common editorial tasks.

For a list of things I can do to help you, just type:

@editorialbot commands

For example, to regenerate the paper pdf after making changes in the paper's md or bib files, type:

@editorialbot generate pdf

[editorialbot]

Software report:

github.com/AlDanial/cloc v 1.90  T=0.13 s (962.5 files/s, 185496.1 lines/s)
-------------------------------------------------------------------------------
Language                     files          blank        comment           code
-------------------------------------------------------------------------------
Julia                          102           3043           2561           8746
SVG                              2              0            150           7699
Markdown                         6            168              0            427
TeX                              1             29              0            253
YAML                             7             24              7            143
TOML                             2              5              0             32
Lisp                             1              8              0             25
-------------------------------------------------------------------------------
SUM:                           121           3277           2718          17325
-------------------------------------------------------------------------------

Commit count by author:

    93  philippwitte
    61  rafaelorozco
    53  Gabrio Rizzuti
    48  Philipp Witte
    34  Mathias Louboutin
    28  Rafael
    23  mloubout
    22  rafael orozco
    21  Ali Siahkoohi
     9  Rafael Orozco
     5  CompatHelper Julia
     5  Orozco
     4  Felix Herrmann
     4  pwitte
     3  gabrio
     3  pwitte3
     2  PetersBas
     2  Páll Haraldsson
     2  felix
     1  Ziyi (Francis) Yin
     1  alisiahkoohi

[editorialbot]

Paper file info:

📄 Wordcount for paper.md is 1097

✅ The paper includes a Statement of need section

[editorialbot]

License info:

✅ License found: MIT License (Valid open source OSI approved license)

[editorialbot]

:point_right::page_facing_up: Download article proof :page_facing_up: View article proof on GitHub :page_facing_up: :point_left:

[editorialbot]

Reference check summary (note 'MISSING' DOIs are suggestions that need verification):

OK DOIs

- 10.5281/zenodo.10100624 is OK
- 10.5281/zenodo.4296287 is OK

MISSING DOIs

- 10.23952/jano.4.2022.2.05 may be a valid DOI for title: Point-TO-SET DISTANCE FUNCTIONS FOR OUTPUT-CONSTRA...
- No DOI given, and none found for title: Julia: A fast dynamic language for technical compu...
- No DOI given, and none found for title: Symmetric block-low-rank layers for fully reversib...
- No DOI given, and none found for title: Memory Efficient Invertible Neural Networks for 3D...
- 10.1109/tnnls.2020.3042395 may be a valid DOI for title: BayesFlow: Learning complex stochastic models with...
- No DOI given, and none found for title: A differentiable programming system to bridge mach...
- No DOI given, and none found for title: Framework for Easily Invertible Architectures (FrE...
- No DOI given, and none found for title: Automatic differentiation in pytorch
- 10.1007/bf01456927 may be a valid DOI for title: Zur theorie der orthogonalen funktionensysteme
- No DOI given, and none found for title: Nice: Non-linear independent components estimation
- No DOI given, and none found for title: Density estimation using real nvp
- No DOI given, and none found for title: Fully hyperbolic convolutional neural networks
- 10.1609/aaai.v35i9.16997 may be a valid DOI for title: HINT: Hierarchical invertible neural transport for...
- No DOI given, and none found for title: Glow: Generative flow with invertible 1x1 convolut...
- 10.21105/joss.05361 may be a valid DOI for title: normflows: A PyTorch Package for Normalizing Flows
- No DOI given, and none found for title: Enabling uncertainty quantification for seismic da...
- 10.1190/segam2020-3428150.1 may be a valid DOI for title: Parameterizing uncertainty by deep invertible netw...
- No DOI given, and none found for title: Preconditioned training of normalizing flows for v...
- No DOI given, and none found for title: Wave-equation-based inversion with amortized varia...
- No DOI given, and none found for title: Refining Amortized Posterior Approximations using ...
- 10.1190/tle42070474.1 may be a valid DOI for title: Learned multiphysics inversion with differentiable...
- No DOI given, and none found for title: Photoacoustic imaging with conditional priors from...
- 10.1190/geo2022-0472.1 may be a valid DOI for title: Reliable amortized variational inference with phys...
- 10.1186/s40323-023-00252-0 may be a valid DOI for title: Solving multiphysics-based inverse problems with l...
- No DOI given, and none found for title: Inference of CO2 flow patterns–a feasibility study
- No DOI given, and none found for title: Amortized Normalizing Flows for Transcranial Ultra...
- 10.1117/12.2651691 may be a valid DOI for title: Adjoint operators enable fast and amortized machin...
- 10.1109/tci.2023.3248949 may be a valid DOI for title: Conditional injective flows for Bayesian imaging

INVALID DOIs

- None

[drvinceknight]

👋🏼 @rafaelorozco @aurorarossi & @Nando-Hegemann this is the review thread for the paper. All of our communications will happen here from now on.

As a reviewer, the first step is to create a checklist for your review by entering

@editorialbot generate my checklist

as the top of a new comment in this thread.

These checklists contain the JOSS requirements. As you go over the submission, please check any items that you feel have been satisfied. The first comment in this thread also contains links to the JOSS reviewer guidelines.

The JOSS review is different from most other journals. Our goal is to work with the authors to help them meet our criteria instead of merely passing judgment on the submission. As such, the reviewers are encouraged to submit issues and pull requests on the software repository. When doing so, please mention openjournals/joss-reviews#6554 so that a link is created to this thread (and I can keep an eye on what is happening). Please also feel free to comment and ask questions on this thread. In my experience, it is better to post comments/questions/suggestions as you come across them instead of waiting until you've reviewed the entire package.

We aim for reviews to be completed within about 2-4 weeks. Please let me know if any of you require some more time. We can also use EditorialBot (our bot) to set automatic reminders if you know you'll be away for a known period of time.

Please feel free to ping me (@drvinceknight ) if you have any questions/concerns.

[aurorarossi]

Review checklist for @aurorarossi

Conflict of interest

• [x] I confirm that I have read the JOSS conflict of interest (COI) policy and that: I have no COIs with reviewing this work or that any perceived COIs have been waived by JOSS for the purpose of this review.

Code of Conduct

• [x] I confirm that I read and will adhere to the JOSS code of conduct.

General checks

• [x] Repository: Is the source code for this software available at the https://github.com/slimgroup/InvertibleNetworks.jl?
• [x] License: Does the repository contain a plain-text LICENSE or COPYING file with the contents of an OSI approved software license?
• [x] Contribution and authorship: Has the submitting author (@rafaelorozco) made major contributions to the software? Does the full list of paper authors seem appropriate and complete?
• [x] Substantial scholarly effort: Does this submission meet the scope eligibility described in the JOSS guidelines
• [x] Data sharing: If the paper contains original data, data are accessible to the reviewers. If the paper contains no original data, please check this item.
• [x] Reproducibility: If the paper contains original results, results are entirely reproducible by reviewers. If the paper contains no original results, please check this item.
• [x] Human and animal research: If the paper contains original data research on humans subjects or animals, does it comply with JOSS's human participants research policy and/or animal research policy? If the paper contains no such data, please check this item.

Functionality

• [x] Installation: Does installation proceed as outlined in the documentation?
• [x] Functionality: Have the functional claims of the software been confirmed?
• [x] Performance: If there are any performance claims of the software, have they been confirmed? (If there are no claims, please check off this item.)

Documentation

• [x] A statement of need: Do the authors clearly state what problems the software is designed to solve and who the target audience is?
• [x] Installation instructions: Is there a clearly-stated list of dependencies? Ideally these should be handled with an automated package management solution.
• [x] Example usage: Do the authors include examples of how to use the software (ideally to solve real-world analysis problems).
• [x] Functionality documentation: Is the core functionality of the software documented to a satisfactory level (e.g., API method documentation)?
• [x] Automated tests: Are there automated tests or manual steps described so that the functionality of the software can be verified?
• [x] Community guidelines: Are there clear guidelines for third parties wishing to 1) Contribute to the software 2) Report issues or problems with the software 3) Seek support

Software paper

• [x] Summary: Has a clear description of the high-level functionality and purpose of the software for a diverse, non-specialist audience been provided?
• [x] A statement of need: Does the paper have a section titled 'Statement of need' that clearly states what problems the software is designed to solve, who the target audience is, and its relation to other work?
• [x] State of the field: Do the authors describe how this software compares to other commonly-used packages?
• [x] Quality of writing: Is the paper well written (i.e., it does not require editing for structure, language, or writing quality)?
• [x] References: Is the list of references complete, and is everything cited appropriately that should be cited (e.g., papers, datasets, software)? Do references in the text use the proper citation syntax?

[aurorarossi]

@rafaelorozco There are many examples in your repository, so it would be nice if you could also add the code to reproduce the benchmark plots of the paper using Pytorch.

[Nando-Hegemann]

Review checklist for @Nando-Hegemann

Conflict of interest

• [x] I confirm that I have read the JOSS conflict of interest (COI) policy and that: I have no COIs with reviewing this work or that any perceived COIs have been waived by JOSS for the purpose of this review.

Code of Conduct

• [x] I confirm that I read and will adhere to the JOSS code of conduct.

General checks

• [x] Repository: Is the source code for this software available at the https://github.com/slimgroup/InvertibleNetworks.jl?
• [x] License: Does the repository contain a plain-text LICENSE or COPYING file with the contents of an OSI approved software license?
• [x] Contribution and authorship: Has the submitting author (@rafaelorozco) made major contributions to the software? Does the full list of paper authors seem appropriate and complete?
• [x] Substantial scholarly effort: Does this submission meet the scope eligibility described in the JOSS guidelines
• [x] Data sharing: If the paper contains original data, data are accessible to the reviewers. If the paper contains no original data, please check this item.
• [ ] Reproducibility: If the paper contains original results, results are entirely reproducible by reviewers. If the paper contains no original results, please check this item.
• [x] Human and animal research: If the paper contains original data research on humans subjects or animals, does it comply with JOSS's human participants research policy and/or animal research policy? If the paper contains no such data, please check this item.

Functionality

• [x] Installation: Does installation proceed as outlined in the documentation?
• [ ] Functionality: Have the functional claims of the software been confirmed?
• [ ] Performance: If there are any performance claims of the software, have they been confirmed? (If there are no claims, please check off this item.)

Documentation

• [x] A statement of need: Do the authors clearly state what problems the software is designed to solve and who the target audience is?
• [x] Installation instructions: Is there a clearly-stated list of dependencies? Ideally these should be handled with an automated package management solution.
• [ ] Example usage: Do the authors include examples of how to use the software (ideally to solve real-world analysis problems).
• [x] Functionality documentation: Is the core functionality of the software documented to a satisfactory level (e.g., API method documentation)?
• [x] Automated tests: Are there automated tests or manual steps described so that the functionality of the software can be verified?
• [ ] Community guidelines: Are there clear guidelines for third parties wishing to 1) Contribute to the software 2) Report issues or problems with the software 3) Seek support

Software paper

• [x] Summary: Has a clear description of the high-level functionality and purpose of the software for a diverse, non-specialist audience been provided?
• [x] A statement of need: Does the paper have a section titled 'Statement of need' that clearly states what problems the software is designed to solve, who the target audience is, and its relation to other work?
• [ ] State of the field: Do the authors describe how this software compares to other commonly-used packages?
• [ ] Quality of writing: Is the paper well written (i.e., it does not require editing for structure, language, or writing quality)?
• [x] References: Is the list of references complete, and is everything cited appropriately that should be cited (e.g., papers, datasets, software)? Do references in the text use the proper citation syntax?

[Nando-Hegemann]

I have only a few minor remarks:

General Checks -> Reproducibility: I'm not able to reproduce the results of the paper (or run any of the other examples). I'm aware that this is due to my lack of familiarity with the Julia language, but maybe you could include a simple section on "How to run the MNIST example" (simple command line instructions) in the README/Doc as well as a description on how to set up an environment to execute the examples/tests. Right now, after freshly installing Julia and running the scripts/tests I simply get LoadError(s).

Functionality -> Functionality & Functionality -> Performance: I agree with @aurorarossi that a script reproducing the plots of the paper would be good. A simple (python) script and respective environment for the python setup used to produce the results would be a good addition.

Documentation -> Example usage:

1. You have a lot of example scripts of which only a few are included in the online documentation (which is fine). However, the examples in the online documentation on provide the source code of the scripts and an image. I think providing more context to the examples (what is the setting, why is it interesting as an example) and describing how the individual steps work and what the output picture depicts would be good.
2. Currently you copy-pasted source code from your scripts to the examples in the doc. To decrease maintenance issues, I'd suggest referencing (certain lines of) the Julia scripts in the doc, so that the doc always shows source code from executable files. This way your doc and the example scripts don't run out of sync if you change something.

With .rst files you could do something like

.. literalinclude:: ../examples/applications/file_name.jl
  :language: julia
  :lines: 8-9

Documentation -> Community Guidelines: Again, I agree with @aurorarossi to include a clearly visible link to your general contribution guidelines in your README.

Software Paper -> State of the field: In the paper you describe that your software performs faster then other packages such as FrEIA or normflows, but I'd suggest adding some additional notes on what separates this software from other packages. For example you could focus on the following:

1. Is there a reason to implement Invertible Networks in Julia rather than in python?
2. Does your software include algorithms/architectures/training routines not implemented anywhere else?
3. Does your software combine established algorithms from multiple other packages, thus providing a more versatile and easy to use framework? In other words, you should reason why you implemented a new software package. If you mainly compute gradients analytically instead of using autograd one could argue that it would be better to integrate these computational improvements in other already existing packages. Also, are there any other Julia packages for INNs? If so they should be mentioned and differences should be explained as well.

Software Paper -> Quality of writing: I have only some minor formulation changes and suggestions. See https://github.com/slimgroup/InvertibleNetworks.jl/pull/108.

[rafaelorozco]

Hello @Nando-Hegemann thank you for your useful comments and feedback! I will be integrating them over the course of the next few days.

For anyone reading this thread, I just want to clarify that we never claim to this software performs faster than non-invertiblity-exploiting frameworks. In fact, due to recomputing the intermediate activations this package should use more compute time. The main bottleneck in normalizing flows has been memory efficiency (cited in the paper) so this is a fair price to pay for enabling inference on large inputs and deep architectures.

[danielskatz]

👋 @rafaelorozco - I just wanted to check on how your work on this is going. (As track editor, I try to check on reviews in the CSISM track where no progress has been recorded in a 2-week period.)

[rafaelorozco]

Hello things are going well, I have been chipping away at the recommendations made by the reviewers.

I added the contribution guidelines as requested.

I added some scripts to reproduce the benchmarks in the manuscript https://github.com/slimgroup/InvertibleNetworks.jl/pull/109

I also changed some of the dependencies so that it is easier to install the package. @Nando-Hegemann was not able to install the package although I dont know what the error was so I took away pyplots as a dependency from the first example which might cause problems. If you are still getting errors please let me know what they are.

[rafaelorozco]

@editorialbot check references

[editorialbot]

Reference check summary (note 'MISSING' DOIs are suggestions that need verification):

OK DOIs

- 10.23952/jano.4.2022.2.05 is OK
- 10.1109/tnnls.2020.3042395 is OK
- 10.5281/zenodo.10100624 is OK
- 10.5281/zenodo.4296287 is OK
- 10.1007/bf01456927 is OK
- 10.1609/aaai.v35i9.16997 is OK
- 10.21105/joss.05361 is OK
- 10.1190/segam2020-3428150.1 is OK
- 10.1190/tle42070474.1 is OK
- 10.1190/geo2022-0472.1 is OK
- 10.1186/s40323-023-00252-0 is OK
- 10.1117/12.2651691 is OK
- 10.1109/tci.2023.3248949 is OK

MISSING DOIs

- 10.1190/image2022-3750561.1 may be a valid DOI for title: Accelerating innovation with software abstractions...
- No DOI given, and none found for title: NormalizingFlows.jl
- No DOI given, and none found for title: Bijectors. jl: Flexible transformations for probab...
- 10.52591/lxai202312101 may be a valid DOI for title: Self-consuming generative models go mad
- 10.1137/141000671 may be a valid DOI for title: Julia: A fresh approach to numerical computing
- No DOI given, and none found for title: Symmetric block-low-rank layers for fully reversib...
- No DOI given, and none found for title: Memory Efficient Invertible Neural Networks for 3D...
- No DOI given, and none found for title: A differentiable programming system to bridge mach...
- No DOI given, and none found for title: Framework for Easily Invertible Architectures (FrE...
- No DOI given, and none found for title: Automatic differentiation in pytorch
- No DOI given, and none found for title: Nice: Non-linear independent components estimation
- No DOI given, and none found for title: Density estimation using real nvp
- No DOI given, and none found for title: Fully hyperbolic convolutional neural networks
- No DOI given, and none found for title: Glow: Generative flow with invertible 1x1 convolut...
- No DOI given, and none found for title: Enabling uncertainty quantification for seismic da...
- No DOI given, and none found for title: Preconditioned training of normalizing flows for v...
- No DOI given, and none found for title: Wave-equation-based inversion with amortized varia...
- No DOI given, and none found for title: Refining Amortized Posterior Approximations using ...
- No DOI given, and none found for title: Photoacoustic imaging with conditional priors from...
- No DOI given, and none found for title: Inference of CO2 flow patterns–a feasibility study
- No DOI given, and none found for title: Amortized Normalizing Flows for Transcranial Ultra...

INVALID DOIs

- None

[rafaelorozco]

@editorialbot check references

[editorialbot]

Reference check summary (note 'MISSING' DOIs are suggestions that need verification):

OK DOIs

- 10.1190/image2022-3750561.1 is OK
- 10.52591/lxai202312101 is OK
- 10.23952/jano.4.2022.2.05 is OK
- 10.1137/141000671 is OK
- 10.48550/arXiv.1912.12137 is OK
- 10.48550/arXiv.2204.11850 is OK
- 10.1109/tnnls.2020.3042395 is OK
- 10.48550/arXiv.1907.07587 is OK
- 10.5281/zenodo.10100624 is OK
- 10.5281/zenodo.4296287 is OK
- 10.1007/bf01456927 is OK
- 10.48550/arXiv.1410.8516 is OK
- 10.48550/arXiv.1605.08803 is OK
- 10.1007/s40687-022-00343-1 is OK
- 10.1609/aaai.v35i9.16997 is OK
- 10.21105/joss.05361 is OK
- 10.1190/segam2021-3583705.1 is OK
- 10.1190/segam2020-3428150.1 is OK
- 10.48550/arXiv.2101.03709 is OK
- 10.48550/arXiv.2203.15881 is OK
- 10.48550/arXiv.2305.08733 is OK
- 10.1190/tle42070474.1 is OK
- 10.1190/geo2022-0472.1 is OK
- 10.1186/s40323-023-00252-0 is OK
- 10.48550/arXiv.2311.00290 is OK
- 10.48550/arXiv.2303.03478 is OK
- 10.1117/12.2651691 is OK
- 10.1109/tci.2023.3248949 is OK

MISSING DOIs

- No DOI given, and none found for title: NormalizingFlows.jl
- No DOI given, and none found for title: Bijectors. jl: Flexible transformations for probab...
- No DOI given, and none found for title: Framework for Easily Invertible Architectures (FrE...
- No DOI given, and none found for title: Glow: Generative flow with invertible 1x1 convolut...
- No DOI given, and none found for title: Photoacoustic imaging with conditional priors from...

INVALID DOIs

- doi.org/10.48550/arXiv.1912.01703 is INVALID because of 'doi.org/' prefix

[rafaelorozco]

@editorialbot generate pdf

[editorialbot]

:point_right::page_facing_up: Download article proof :page_facing_up: View article proof on GitHub :page_facing_up: :point_left:

[rafaelorozco]

Hello all, here are my responses and integrations of the comments from @Nando-Hegemann

General Checks -> Reproducibility: I'm not able to reproduce the results of the paper (or run any of the other examples). I'm aware that this is due to my lack of familiarity with the Julia language, but maybe you could include a simple section on "How to run the MNIST example" (simple command line instructions) in the README/Doc as well as a description on how to set up an environment to execute the examples/tests. Right now, after freshly installing Julia and running the scripts/tests I simply get LoadError(s).

It was not clear what the error you are refering to was so I tried running the example from various systems and arrived to the conclusion that the pyplots dependency might be an issue if the system python does not have matplotlib installed. To resolve this I took away that dependency from the example and instead use Plots.jl. I also add lines to manually add the package dependencies in case new Julia users do not know how to add those.

Functionality -> Functionality & Functionality -> Performance: I agree with @aurorarossi that a script reproducing the plots of the paper would be good. A simple (python) script and respective environment for the python setup used to produce the results would be a good addition.

Great idea! I added the benchmark scripts in this PR https://github.com/slimgroup/InvertibleNetworks.jl/pull/109

Documentation -> Example usage:

1. You have a lot of example scripts of which only a few are included in the online documentation (which is fine). However, the examples in the online documentation on provide the source code of the scripts and an image. I think providing more context to the examples (what is the setting, why is it interesting as an example) and describing how the individual steps work and what the output picture depicts would be good.

Thank you for the feedback, I completely agree and I added some more comments to the examples in the online documentation

2. Currently you copy-pasted source code from your scripts to the examples in the doc. To decrease maintenance issues, I'd suggest referencing (certain lines of) the Julia scripts in the doc, so that the doc always shows source code from executable files. This way your doc and the example scripts don't run out of sync if you change something.

With .rst files you could do something like

.. literalinclude:: ../examples/applications/file_name.jl
  :language: julia
  :lines: 8-9

This is a very useful suggestion. Currently those example are being run as code during documentation generation and I was unable to get that working in tandem with the .rst files. But I will keep an eye out in the future to see if we can integrate both of these methods.

Documentation -> Community Guidelines: Again, I agree with @aurorarossi to include a clearly visible link to your general contribution guidelines in your README.

Thank you for the feedback we have added some guidelines.

Software Paper -> State of the field: In the paper you describe that your software performs faster then other packages such as FrEIA or normflows, but I'd suggest adding some additional notes on what separates this software from other packages. For example you could focus on the following:

1. Is there a reason to implement Invertible Networks in Julia rather than in python?

From the onset, our goal was interoperability of this package with other packages in the workflow of our lab. The multiple dispatch system of Julia as a fundamental design feature greatly aided this goal for reference please see this publication https://arxiv.org/pdf/2203.15038. Other than this programatic reason, we are interested in scalable software since our goal is to offer solutions for imaging problems that have high degrees of freedom and require efficient solutions. I have added some sentences to describe these reasons.

2. Does your software include algorithms/architectures/training routines not implemented anywhere else?
3. Does your software combine established algorithms from multiple other packages, thus providing a more versatile and easy to use framework?

Here we are only claiming to uniquely have access to memory efficiency that has enabled its application in large scale imaging problems. This has been a huge bottleneck for the acceptance of normalizing flows so we believe it is important for the community.

In other words, you should reason why you implemented a new software package. If you mainly compute gradients analytically instead of using autograd one could argue that it would be better to integrate these computational improvements in other already existing packages.

Im afraid it is a bit difficult to phrase these arguments since this package was implemented 4 years ago when there were few options for existing INN packages. It is now has mature functionality that has been used for many publications.

Also, are there any other Julia packages for INNs? If so they should be mentioned and differences should be explained as well.

Some packages that come to mind are: Bijectors.jl, NormalizingFlows.jl. The important difference being that they do not implement the manually defined gradients required for memory efficiency of large scale training. I have added them to the writeup.

Software Paper -> Quality of writing: I have only some minor formulation changes and suggestions. See slimgroup/InvertibleNetworks.jl#108.

Thank you for the suggestions, I have incorporated them into the writeup.

[rafaelorozco]

@drvinceknight I believe that I have responded to all of the remarks from the two reviewers and have integrated these as best as possible into the package and writeup. I hope that this round of revision is agreeable to the reviewers and will be happy to iterate further!

[aurorarossi]

The revisions are satisfactory to me.

[rafaelorozco]

@drvinceknight how should we proceed from here? Thank you for the continuing help.

[drvinceknight]

This is great, thanks to both of the reviewers.

I'll take things from here @rafaelorozco, I need to do a few checks and will get back to you (there might be one or two minor things that are needed). I'm a bit busy today but will do it tomorrow.

[drvinceknight]

Post-Review Checklist for Editor and Authors

Additional Author Tasks After Review is Complete

• Double check authors and affiliations (including ORCIDs)
• Make a release of the software with the latest changes from the review and post the version number here. This is the version that will be used in the JOSS paper.
• Archive the release on Zenodo/figshare/etc and post the DOI here.
• Make sure that the title and author list (including ORCIDs) in the archive match those in the JOSS paper.
• Make sure that the license listed for the archive is the same as the software license.

Editor Tasks Prior to Acceptance

• [x] Read the text of the paper and offer comments/corrections (as either a list or a pull request)
• [x] Check that the archive title, author list, version tag, and the license are correct
• [x] Set archive DOI with @editorialbot set <DOI here> as archive
• [x] Set version with @editorialbot set <version here> as version
• [x] Double check rendering of paper with @editorialbot generate pdf
• [x] Specifically check the references with @editorialbot check references and ask author(s) to update as needed
• [x] Recommend acceptance with @editorialbot recommend-accept

[drvinceknight]

@editorialbot generate pdf

[editorialbot]

:point_right::page_facing_up: Download article proof :page_facing_up: View article proof on GitHub :page_facing_up: :point_left:

[aurorarossi]

I'm a bit confused, did @Nando-Hegemann check the last changes? if I'm not mistaken I see his last activity in this PR on May 22nd.

[drvinceknight]

I'm a bit confused, did @Nando-Hegemann check the last changes? if I'm not mistaken I see his last activity in this PR on May 22nd.

Thanks @aurorarossi, given the nature of his comments I am satisfied that they have been addressed. @rafaelorozco would you agree?

[drvinceknight]

@rafaelorozco I have made some minor suggestions here: https://github.com/slimgroup/InvertibleNetworks.jl/pull/111

[drvinceknight]

Once you have checked those suggestions I'll need you to create a new archive where the archive title matches the title (of the paper) the author list and the version tag,

[rafaelorozco]

For sure!

The version tag is 2.2.9 and here is the new archive https://zenodo.org/records/12810006 with DOI: 10.5281/zenodo.12810006

[drvinceknight]

@editorialbot set 10.5281/zenodo.12810006 as archive

[editorialbot]

Done! archive is now 10.5281/zenodo.12810006

[drvinceknight]

@editorialbot set 2.2.9 as version

[editorialbot]

Done! version is now 2.2.9

[drvinceknight]

@editorialbot recommend-accept

[editorialbot]

Attempting dry run of processing paper acceptance...

[editorialbot]

Reference check summary (note 'MISSING' DOIs are suggestions that need verification):

OK DOIs

- 10.1190/image2022-3750561.1 is OK
- 10.52591/lxai202312101 is OK
- 10.23952/jano.4.2022.2.05 is OK
- 10.1137/141000671 is OK
- 10.48550/arXiv.1912.12137 is OK
- 10.48550/arXiv.2204.11850 is OK
- 10.1109/tnnls.2020.3042395 is OK
- 10.48550/arXiv.1907.07587 is OK
- 10.5281/zenodo.10100624 is OK
- 10.5281/zenodo.4296287 is OK
- 10.48550/arXiv.1912.01703 is OK
- 10.1007/bf01456927 is OK
- 10.48550/arXiv.1410.8516 is OK
- 10.48550/arXiv.1605.08803 is OK
- 10.1007/s40687-022-00343-1 is OK
- 10.1609/aaai.v35i9.16997 is OK
- 10.21105/joss.05361 is OK
- 10.1190/segam2021-3583705.1 is OK
- 10.1190/segam2020-3428150.1 is OK
- 10.48550/arXiv.2101.03709 is OK
- 10.48550/arXiv.2203.15881 is OK
- 10.48550/arXiv.2305.08733 is OK
- 10.1190/tle42070474.1 is OK
- 10.1190/geo2022-0472.1 is OK
- 10.1186/s40323-023-00252-0 is OK
- 10.48550/arXiv.2311.00290 is OK
- 10.48550/arXiv.2303.03478 is OK
- 10.1117/12.2651691 is OK
- 10.1109/tci.2023.3248949 is OK

MISSING DOIs

- No DOI given, and none found for title: NormalizingFlows.jl
- No DOI given, and none found for title: Bijectors. jl: Flexible transformations for probab...
- No DOI given, and none found for title: Framework for Easily Invertible Architectures (FrE...
- No DOI given, and none found for title: Glow: Generative flow with invertible 1x1 convolut...
- No DOI given, and none found for title: Photoacoustic imaging with conditional priors from...

INVALID DOIs

- None

[editorialbot]

:wave: @openjournals/csism-eics, this paper is ready to be accepted and published.

Check final proof :point_right::page_facing_up: Download article

If the paper PDF and the deposit XML files look good in https://github.com/openjournals/joss-papers/pull/5694, then you can now move forward with accepting the submission by compiling again with the command @editorialbot accept

[danielskatz]

@rafaelorozco - As the track editor, I'll next check and proofread this, and let you know what else, if anything, is needed.

[danielskatz]

@rafaelorozco - I'm suggesting the changes in https://github.com/slimgroup/InvertibleNetworks.jl/pull/113 - please merge this, or let me know what you disagree with, then we can continue the acceptance process

[rafaelorozco]

@danielskatz Merged, thank you for the suggestions!

[danielskatz]

@editorialbot recommend-accept

[editorialbot]

Attempting dry run of processing paper acceptance...

[editorialbot]

Reference check summary (note 'MISSING' DOIs are suggestions that need verification):

OK DOIs

- 10.1190/image2022-3750561.1 is OK
- 10.52591/lxai202312101 is OK
- 10.23952/jano.4.2022.2.05 is OK
- 10.1137/141000671 is OK
- 10.48550/arXiv.1912.12137 is OK
- 10.48550/arXiv.2204.11850 is OK
- 10.1109/tnnls.2020.3042395 is OK
- 10.48550/arXiv.1907.07587 is OK
- 10.5281/zenodo.10100624 is OK
- 10.5281/zenodo.4296287 is OK
- 10.48550/arXiv.1912.01703 is OK
- 10.1007/bf01456927 is OK
- 10.48550/arXiv.1410.8516 is OK
- 10.48550/arXiv.1605.08803 is OK
- 10.1007/s40687-022-00343-1 is OK
- 10.1609/aaai.v35i9.16997 is OK
- 10.21105/joss.05361 is OK
- 10.1190/segam2021-3583705.1 is OK
- 10.1190/segam2020-3428150.1 is OK
- 10.48550/arXiv.2101.03709 is OK
- 10.48550/arXiv.2203.15881 is OK
- 10.48550/arXiv.2305.08733 is OK
- 10.1190/tle42070474.1 is OK
- 10.1190/geo2022-0472.1 is OK
- 10.1186/s40323-023-00252-0 is OK
- 10.48550/arXiv.2311.00290 is OK
- 10.48550/arXiv.2303.03478 is OK
- 10.1117/12.2651691 is OK
- 10.1109/tci.2023.3248949 is OK

MISSING DOIs

- No DOI given, and none found for title: NormalizingFlows.jl
- No DOI given, and none found for title: Bijectors. jl: Flexible transformations for probab...
- No DOI given, and none found for title: Framework for Easily Invertible Architectures (FrE...
- No DOI given, and none found for title: Glow: Generative flow with invertible 1x1 convolut...
- No DOI given, and none found for title: Photoacoustic imaging with conditional priors from...

INVALID DOIs

- None

[editorialbot]

:wave: @openjournals/csism-eics, this paper is ready to be accepted and published.

Check final proof :point_right::page_facing_up: Download article

If the paper PDF and the deposit XML files look good in https://github.com/openjournals/joss-papers/pull/5701, then you can now move forward with accepting the submission by compiling again with the command @editorialbot accept

[danielskatz]

@rafaelorozco - can you confirm that you don't want to have an acknowledgements section, and that there's no funding or anything else you need to acknowledge?

[rafaelorozco]

Sorry, I need to add some acknowledgements. I will push that version as soon as possible.