Terminology

[FabienCharmet]

Which one should we use ?

Explainable techniques, explanation methods ?

I'm not sure.

[PFGimenez]

I prefer explanation methods, as in "methods that can generate explanation". I would not say that LIME is an "explainable technique", for example.

[cloudgravity]

I agree with PF. "Explainable technique" means that the technique is "intrinsically explainable", and can characterize ML methods such as DTs, for example. "Explanation method" characterizes the fact that it enables the explanation of other (non-explainable) techniques. SHAP or LIME are "explanation methods".

[cloudgravity]

I have another terminology issue : "dataset" vs "database". Some of the contributors seem to prefer "database" or use it interchangeably with "dataset". In my understanding (and according to https://www.usgs.gov/faqs/what-are-differences-between-data-dataset-and-database), they mean different things: usually a database is a collection of dataset with means (a system) to access and manipulate the data. While some of the datasets mentioned in the survey could be considered databases, I would still enforce the use of "dataset" most of the times.

[cloudgravity]

Acronyms and capitalization is also another issue of discrepancy between contributors, e.g.:

• when writing "a decision tree", prefer using lowercase; when writing "a Decision Tree classifier", you may use capitals
• "Random Forest" is often abbreviated as "RF"

[cloudgravity]

With respect to spelling issues, we should have a shared glossary but I guess Harry will uniformize everything in the end. For the sake of it, I am starting a short list to be expanded:

• black-box
• dataset
• cybersecurity
• cyber-attacks