[jre]

I don't know about the others, but the two visions dataset they compare to (MNIST and the face recognition one) are small datasets and the CNN they compare to doesn't seem very state of the art.

It also seems each layer of random forest just concatenates a class distribution to the original feature vector. So this doesn't seem to get the same "hierarchy of features" benefit that you get in large-scale CNN and DNN.

[jkbschwarz]

To your point that they are comparing small datasets. I dont see that as a problem. If they achieve better results on small datasets that is a great achievement, as often the bottleneck is the size of the dataset rather than computation time.

[krona]

> often the bottleneck is the size of the dataset rather than computation time

That's generally true for DNNs, which is a good place to be if you have lots of data. This typically isn't true for tree based approaches, which is why they fell out of fashion in some problem domains; they don't generalize as well. This paper doesn't seem to change what we already know in this respect.

[argonaut]

They achieve substantially worse results than state of the art on (the toy dataset) MNIST.

[aerioux]

^ The authors time and effort they observe it takes to create state-of-the-art CNNs, but their point-of-comparison CNNs look to be fairly simple -- I don't see an AlexNet or something for some of these tasks either just as a point of comparison even if not a fully relevant one

[curuinor]

Comparable to SOTA circa 1998.