Closed jaclyn-taroni closed 3 years ago
AppVeyor build 1.0.212 for commit 1030e9fea7866c27a21474928d24cc100f2e5153 is now complete. Found 0 potential spelling error(s). Preview:
...
AppVeyor build 1.0.213 for commit ccf7764aa110312fbbba417ad3b45677932c1a40 is now complete. Found 0 potential spelling error(s). Preview:
...
I too support this proposal. :) :shipit:
I think the first section could use a figure, though might run the risk of veering into "review" territory
Everything we wanted to track is now tracked, so I will go ahead and merge this!
⚠️ This is ready to be looked at and discussed, but, if folks agree with the restructuring plan, it should not be merged until we have a plan for those next steps! ⚠️
Purpose
I thought I was going to quickly address the rewrite of the transfer learning and multi-task learning sections (#110) and then move onto the putting it all together section (#111), but as I was editing
05.prior-knowledge.md
the representation learning section felt out of place. It occurred to me we talk about dimensionality reduction in the first section (03.heterogeneity.md
).I propose that we make the section that is about "Managing disparities in data generation is required for robust analysis" more general – let's talk about issues related to features >> samples, characterizing structure in your data, feature selection and feature learning. By first talking about dimension reduction more generally, we can then focus in on how that can reveal structure that's not of particular scientific interest (batch effects) as a specific motivation or use case. I also think talking about feature selection (the why) in the first section could be a helpful bit of set up for some of the points in
04.model-complexity.md
.To this end, I've done the following in this pull request:
03.heterogeneity.md
(no renaming, yet) such thatDimensionality reduction and representation learning
Visualization and correction of technical effects
05.prior-knowledge.md
to the section of03.heterogeneity.md
that also covers dimensionality reduction05.prior-knowledge.md
to a new file06.multiple-approaches-required.md
Directions for reviewers
The diff for this is not going to be particularly useful. For the most part, no small scale text changes of particular consequence have been made. (You will find that the flow is absolutely bonkers as a result 😀 ) Instead, I thought we should discuss this proposed restructuring and then track and split up work such that I am not blocking everyone else.
As such, please focus on the big picture 🖼️ and if you agree with these changes to the structure in principal. If you do agree, do you think the new first section could use its own figure?
Manuscript checklist
Unless otherwise noted above, this PR will be considered ready for review when all four items have been checked.