# Article on relative research impact of Bayesian modeling (Stan/PyMC3) vs deep learning (PyTorch, TensorFlow, Keras)
**URL:** https://discourse.mc-stan.org/t/article-on-relative-research-impact-of-bayesian-modeling-stan-pymc3-vs-deep-learning-pytorch-tensorflow-keras/25991
**Category:** Publicity
**Created:** [January 15, 2022, 7:52pm UTC](https://discourse.mc-stan.org/t/article-on-relative-research-impact-of-bayesian-modeling-stan-pymc3-vs-deep-learning-pytorch-tensorflow-keras/25991 "2022-01-15T19:52:34Z")
**Posts on this page:** 4
**Page:** 1
### Author:  [@breckbaldwin](https://discourse.mc-stan.org/u/breckbaldwin)
#### Post date: [January 15, 2022, 7:52pm UTC](https://discourse.mc-stan.org/t/article-on-relative-research-impact-of-bayesian-modeling-stan-pymc3-vs-deep-learning-pytorch-tensorflow-keras/25991/1 "2022-01-15T19:52:34Z")
Bayesian grant writers,
I did a citation count-based impact assessment of deep learning vs Bayesian modeling to hopefully put an end to reviewers assuming that deep learning was doing all the worthwhile science/research. This has been a problem in the past with grant reviews.
It is peer-reviewed, reproducible, and designed to establish the impact of Bayesian software via an authoritative reference–citations are facts right? Sarcasm aside, I think being able to cite research impact in a proposal rather than argue it makes a stronger case and it saves space. I have been spending a page or two defending against deep learning lately so perhaps it will strengthen your proposals too.
The artless title is intended to err on the side of being stunningly obvious about the information contained within.
Thanks to all that helped out with its creation.
‘[https://doi.org/10.1002/ail2.62](https://doi.org/10.1002/ail2.62)’
> **[Deep learning does not replace Bayesian modeling: Comparing research use via...](https://onlinelibrary.wiley.com/doi/10.1002/ail2.62)**
>
> The apparent tsunami of deep learning citation counts over a moderate swell of Bayesian modeling looks very different if computer science journals are absent. Looking more closely at scientific subje...
Citation is:
@article{baldwin2021deep,
title={Deep Learning does not Replace Bayesian Modeling: Comparing research use via citation counting},
author={Baldwin, Breck},
journal={Applied AI Letters},
pages={e62},
year={2021},
publisher={Wiley Online Library}
}
---
### Author:  [@s.maskell](https://discourse.mc-stan.org/u/s.maskell)
#### Post date: [July 19, 2022, 3:55pm UTC](https://discourse.mc-stan.org/t/article-on-relative-research-impact-of-bayesian-modeling-stan-pymc3-vs-deep-learning-pytorch-tensorflow-keras/25991/2 "2022-07-19T15:55:53Z")
I didn’t spot this post at the time so apologies for the latency of response, but with my “Associate Editor of Applied AI Letters” hat overtly on, I’d be very keen to advocate the journal (see here: [https://onlinelibrary.wiley.com/journal/26895595](https://onlinelibrary.wiley.com/journal/26895595)) as somewhere for Stan-relevant papers that might not otherwise have a home (e.g. as a result of being practical and pragmatic enough to be useful!). Hopefully, @breckbaldwin can agree that it was a painless process to submit the paper and that the review process was fair, constructive and reasonably speedy.
Cheers
Simon
---
### Author:  [@breckbaldwin](https://discourse.mc-stan.org/u/breckbaldwin)
#### Post date: [July 19, 2022, 5:06pm UTC](https://discourse.mc-stan.org/t/article-on-relative-research-impact-of-bayesian-modeling-stan-pymc3-vs-deep-learning-pytorch-tensorflow-keras/25991/3 "2022-07-19T17:06:25Z")
The experience with Applied AI Letters was good and fairly quick–few months end-to-end. Reviewers were supportive and high quality, and the editorial feedback was helpful too. The only downside is the publication fee since it is open access so it will cost $2,100 to publish–so consider whether your employer or the like will pay it, mine didn’t want to pay it so I paid it out of pocket.
Breck
---
### Author:  [@s.maskell](https://discourse.mc-stan.org/u/s.maskell)
#### Post date: [July 20, 2022, 6:58am UTC](https://discourse.mc-stan.org/t/article-on-relative-research-impact-of-bayesian-modeling-stan-pymc3-vs-deep-learning-pytorch-tensorflow-keras/25991/4 "2022-07-20T06:58:32Z")
Fair point. It wasn’t applicable in the setting @breckbaldwin was in, but there’s a list of institutions who have an agreement with Wiley that means that there is no (additional) cost to the author for publishing here:
> **[Institutional Payments | Wiley](https://authorservices.wiley.com/author-resources/Journal-Authors/open-access/affiliation-policies-payments/institutional-funder-payments.html)**
>
> Many institutions have agreements with Wiley which allow them to help affiliated authors with open access article publication charges (APCs). These agreements are known as Wiley Open Access Accounts.