search

Benjamin-Lee / deep-rules

Ten Quick Tips for Deep Learning in Biology
https://benjamin-lee.github.io/deep-rules/
Other
227 stars 45 forks source link

Deep learning in bioinformatics: introduction, application, and perspective in big data era #154

Open agitter opened 5 years ago

agitter commented 5 years ago

https://doi.org/10.1101/563601

This is a (yet another) review article that may be of interest because they provide eight examples of representative deep learning problems in biology: https://github.com/lykaust15/Deep_learning_examples It doesn't look like anyone plans to work on #88 so we could consider linking to some hands on references instead like this and #59.

Benjamin-Lee commented 5 years ago

I like how they have a bunch of different architectures and problems represented in their repo. Their notebooks are less than ideal, but I would prefer to avoid the scope creep #88 would entail.

rasbt commented 5 years ago

I think #88 is a bit too ambitious and given that this is a broad audience, it will only be useful to a very very small subset because demos are very problem domain-, dataset-, and tool-specific (also it will become obsolete/become non-working very soon because tools are evolving quickly). What we could do though is make a nice flowchart though for illustrating the main steps and link them to the take-away messages from each tip.

liyu95 commented 5 years ago

Hi, I just found this interesting project by chance and thank you very much for your interest in our review paper!

The idea of writing an article collaboratively is very interesting to me and I am also very interested in contributing to this project if I can make any substantial contributions.

I have the following questions considering this project:

  1. I understand this project works based on pull request but I am a little curious about how you determine the overall structure of the article and whether a section or a paragraph should be included in the final manuscript. I guess the candidate main contents of this article are here: https://github.com/Benjamin-Lee/deep-rules/blob/master/tips.md. But what about the other sections, such as Introduction and Conclusion?
  2. Do you need to draw any figures and flowcharts for this article? The plain text description is not vivid to the readers.
  3. Do you have any deadline and timeline for the submission?

I would appreciate it if you could address my above concerns. Let me know if I need to open an issue for the above questions.

agitter commented 5 years ago

Welcome @lykaust15. The GitHub issues track TODO items and other ideas for updating the manuscript. Issue https://github.com/Benjamin-Lee/deep-rules/issues/116 has some discussion of the intended timeline. There are also open pull requests. Anyone is welcome to comment on or review a pull request.

tips.md is only an outline, and we now have drafts of most of the tips. The actual content is in the numbered .md files in https://github.com/Benjamin-Lee/deep-rules/tree/master/content. There is one for each tip, an intro, conclusion, etc. The Manubot system builds these files every time a pull request is merged, automatically updating the HTML output at https://benjamin-lee.github.io/deep-rules/

Some figures may be a nice addition. I'm not sure how many the article format allows. We have one cartoon about overfitting currently. You could check the issues to see if there are figure ideas you'd like to work on or create a new issue to propose a figure. Either way, I suggest discussing the idea in an issue before you spend time working on an illustration.

liyu95 commented 5 years ago

@agitter Got it! Thank you very much! I will review the current draft, the issues, and the pull requests, seeing the chance of contributing to this project.

Benjamin-Lee commented 5 years ago

Hi @lykaust15! I'm a fan of DeepSimulator (very useful since I don't have a nanopore sequencer–yet) and happy to have you onboard. Sorry for taking so long to reply; I was swamped with exams until yesterday.

@agitter has covered most of the points, but I'll try to clarify some more. With respect to the timeline, Hofstadter's law has struck again. We're behind schedule (🙍‍♂️) but still making progress. We don't have any official deadlines, and I do recognize that people have other research and jobs and lives to attend to. I try to work on this project every day (not expecting anyone else to at all though) to ensure we keep things moving.

The best ways to help are (in no particular order):

  1. Find mistakes in the content, grammar, citations, etc.
  2. Review the current pull requests
  3. Write (via pull requests) anything you think is missing. The issues list has a bunch of topics that people have raised, so that might be a good place to look for things to add. I granted you collaborator status on the project, so feel free to close issues that have been adequately addressed that we missed in #116.
  4. If you're unsure about something, make an issue so we can discuss.
  5. Figures would be very helpful. I don't have any ideas for figures in mind myself (or else I would have made them!) but would certainly welcome ones that help visually explain our ideas.

Welcome to the project!

liyu95 commented 5 years ago

@Benjamin-Lee Thank you very much for such a warm welcome! I will try my best to contribute to the project!