Yes, I thought I’d talk about basic stats and how to evaluate studies because you assumed the study would be flawed. So I went through what makes a flawed study and what doesn’t — you really can examine just the effect of colour. (I love the logic of, you explained basic stats therefore you can’t know anything else.)
Everything I wrote was accurate, you can fact-check it. Whereas you, “The whole point of machine learning is to outsource feature selection to algorithms so that humans can’t mess it up with their monkey brains” — say things that are untrue. Here let’s just Google it:
“ Feature Selection is the process where you automatically or manually select those features which contribute most to your prediction variable or output in which you are interested in. Having irrelevant features in your data can decrease the accuracy of the models and make your model learn based on irrelevant features.” (Hence my point about dimensionality.)”
Feature selection is an early stage of ML (usually just based on which variables are most correlated with the dependent variable, nothing particularly fancy), after which you build models with diff. algorithms to see which performs the best.
You can, BTW, use your monkey brain for feature selection, the proof’s the pudding, if your model performs well on fresh data… it’s a good model. ML isn’t for feature selection, it’s for mining patterns that are the constellation of all these relevant variable values to train your model to predict e.g. attractive vs. not attractive, using different algorithms, some are robust to missing data, some perform better on numeric data etc. (Hence my point about it ALWAYS being a combination of IVs in ML.)
This is all Google-able; fact checking is easy. You read in depth about ML, and you read the academic articles you’re criticising. You don’t need appeals to authority, or for me to list RapidMiner, R and SAS Enterprise Miner. I’m not an expert by any means, but my god, you can check to see that everything I say is accurate, whereas this isn’t true for you assertions.