I think there are people who analyze for thesis and master's thesis, and people who analyze it on a regular basis. This is an article for such people. I will show you the folder structure that has been optimized for over 3 years, so I would appreciate it if you could refer to it or give me feedback that it is better to do this.
Click here for 2 years ago: The strongest folder structure I think in analysis --Qiita
For the time being, I am using Windows10, 64bit, git bash, matlab 2015b, but the content of the article has nothing to do with it.
The following is a compilation of the tree.rb. The obvious thing to do is to separate folders by role to prevent files from cluttering.
I think that both python and matlab can be done like this (I usually use matlab), but what about R language? I will write an article when I understand it.
If there is an extension, it indicates a file, and if it has a /, it indicates a folder.
tree.rb
$ tree.rb
project-root
|-- README.md
|-- .gitignore
|-- data/
| |-- data1/ #data save folder 1
| `-- data2/ #data save folder 2
|-- experiments/
| |-- category1/
| `-- category2/
| |-- experiment1.m
| `-- experiment2.m
|-- libs/
| |-- +common/
| | |-- FigureGenerator.m
| | `-- Utility.m
| |-- +la/
| | |-- hessian.m
| | `-- jacobian.m
| |-- tests/
| | `-- runtests.m # tests/unittest/*Test.Put a function that executes m and returns the test result. test-driven-Development becomes possible.
| `-- yaml/
| `-- YAMLMatlab_0.4.3/ #Very useful external library. res/Read the yaml file below
|-- output/
| |-- docs/
| | |-- others/
| | |-- paper/
| | | |-- yyyy-mm-dd-first-paper/
| | | `-- yyyy-mm-dd-second-paper/
| | `-- thesis/
| | |-- bachelor/
| | `-- master/
| |-- refs/
| `-- slides/
| |-- conference/
| | |-- yyyy-mm-dd-conference1/
| | |-- yyyy-mm-dd-conference2/
| | `-- yyyy-mm-dd-conference3/
| |-- defense/
| | |-- bachelar-midterm/
| | |-- bachelar-final/
| | |-- master-midterm/
| | `-- master-final/
| |-- discussion/
| | |-- yyyy-mm-dd-discussion.pptx
| | `-- yyyy-mm-dd-discussion.pptx
| `-- others/
|-- res/ #yaml is libs/yaml/*It is converted to a matlab variable by the following YamlMatlab. Place other programming related files
| |-- const.yaml
| `-- litrconst.yaml
|-- results/ #results to correspond to the folder / file names under experiments/Create folders etc. below
| |-- category1/
| `-- category2/
| |-- experiment1/
| `-- experiment2/
| |-- data1/
| `-- data2/
|-- scripts/ #A folder that stores common processes that are called from all script files.
`-- tests/
|-- implementation/ #Playground folder for implementation
`-- unittest/ #The folder where the test suite of unittest is stored.
|-- HelperTest.m # helper class' test class
`-- LATest.m # linear algebra functions' test class
experiments, scripts, results, libs, data, res, tests are the folders related to programming. The roles are shown in the table below.
| folder | description |
|---|---|
| experiments | Put a lot of verification scripts |
| scripts | Common script |
| libs | Put a lot of function files |
| data | Keep the original data used for analysis |
| results | A folder to save the analysis results. It is good to have a folder structure that corresponds to experiments |
| res | Put things that you use for programming but not source code, such as yaml files and other config files |
| tests | I have a lot of test classes |
I don't put a program, but I separate important folders for research as output.
| folder | description |
|---|---|
| output | Research output. I wonder if it can be roughly divided into slide, docs, and refs. Place presentation slides on slides, and dissertations, graduation thesis, and master's thesis on docs. |
Please do not use Japanese for the folder name ** absolutely **. I think he is a person who cannot program.
Since the above file structure is all, I will not write a detailed explanation. Please read. If you have any questions, please leave a comment.
It seems that it has been reread a little recently, so I will add it. I think the following Python template generator is also effective. Please use it. CookieCutter for DataScience in Python
I haven't tried it yet so I can't comment at all, but it might look like rails, django in data science.
Recommended Posts