@enroweb this looks really odd... Kernel methods under dimensionality reduction, feature selection under classifiers, neural networks and deep learning as distinct specialties, training data under deep learning, supervised learning under data mining?
We all have different ways of categorizing methods/questions and can bicker about choices for hours, but this is quite unusual. @Dorialexander @joelgombin
@cazencott Thank you for your feedback! What do you think @ejsmdubois?
@enroweb @ejsmdubois I have another question! There's plenty of publications in genetics and bioinformatics journals that propose new AI/ML/stats methods, and are written by people who are considered in the community to be doing research *on* AI. However it seems that in your paper, the venue would be enough to decide that this is research *with* AI, am I correct? Maybe the separation between AI authors and disciplinary authors isn't that strong?
@cazencott @enroweb @ejsmdubois yeah I don't understand how the categories are chosen either
@joelgombin from the paper, they built a list of AI-related keywords, created a data base of papers from the Microsoft Academic Graph restricted to those with the keywords in their title/abstract, looked at keyword co-occurrence in abstracts, built a keyword co-occurrence graph, and applied a community detection algorithm.
@joelgombin
Unfortunately, when checking this I have now read the sentence "Classifiers and dimensionality reduction are strictly related, being indeed the latter a particular form of classification" and that's just plain wrong.
@cazencott @enroweb @ejsmdubois indeed... Thanks for the explanation
Je veux pouvoir acheter un poster de cette dataviz @ejsmdubois !