Sensitivity analysis tutorial #11
Conversation
- First tutorial (in progress): Sensitivity matrix for DC resistivity problem.
- First tutorial (in progress): Sensitivity matrix for DC resistivity.
A few editions in the abstract.
Was using levels.sort, should be sort(levels) to avoid changing the array levels.
|
Try to be consistent with capitalizing letters. Capitalize all significant words. Transitional words like 'a' 'the' 'for' are not capitalized. See https://www.scribbr.com/academic-writing/capitalization-titles-headings/ |
|
Can we get a fancier image than this? |
|
General comments on file storage and structure: Please house notebooks related to sensitivity and resolution analysis in a directory notebooks/13-sensitivity-resolution. As of right now, the name you gave to the notebook seems fine. Data files that are downloaded and used in the tutorials should be housed in an appropriate directory within the assets folder. You have created a folder notebooks/xx-dc2d/century where everything is unzipped. To fix this, you will need to:
In the near future, we will use something called Pooch to download. |
|
Is this example specific to DC resistivity? Or did you want to extent it to a generate concept? If it's DC resistivity specifically, it could be a tutorial within that section of the website. |
|
Great. Thanks for the comments, I'll work on your suggestions. For the picture to be replaced, do you have any suggestions for a new one? I will make a selection of a few alternatives and show you later. |
|
Did we need to define this function? Can't we use once of the functions within SimPEG.utils.io_utils? |
In fact, the utility might also generate the data object automatically. |
Not sure. I'll look at the possibilities within SimPEG. I'm not really happy with this too. |
|
I feel like the content can really be split into 2 parts. In part 1, we use the sensitivities to see how sensitive our survey geometry is to a particular model. In part 2, we use the sensitivities to determine which features we have recovered with a certain degree of confidence. Is that true? If so, we should be more explicit in stating this. |
In the end, we will always need to use some specific method to create the examples, otherwise the tutorial would only have conceptual information, the mathematics and so on. So, I don't know yet what would be the best strategy to organize these tutorials. One option is, as you suggested, to make this one really specific to DC problems, put it in the DC group and write other similar tutorials for the other methods. |
I agree, we can split the tutorial in two parts, making them shorter. One tutorial would focus on the detectability, and the other on the resolution. |
Added a section about uncertainty quantification and resolution.
I've created the first tutorial, about the use of the sensitivity matrix with a few examples.
There are still some gaps that I'm going to fill (e.g in the section describing some theoretical aspects).
The plan is to continue developing more tutorials in this section showing techniques for uncertainty assessment in inverse problems using SimPEG.