4 ms·
> Chances are a lot of machine learning implementations can be replaced with just "boring" statistical models and achieve more power. This is of course true, b
by version_five 4y ago
> Chances are a lot of machine learning implementations can be replaced with just "boring" statistical models and achieve more power.
This is of course true, but the innovation in ML is not that a neural network (or whatever model) is equivalent to something else, it's finding the weights in the first place.