I've been a peer reviewer before, albeit not in the machine learning space. My comments above were general questions and not a review, since I'm now curious about applying this technique in my own work.
When I do review a paper, my guide star is "does the paper answer its own question using a methodology powerful enough to detect if the answer is 'no'?" A secondary question is "are any arbitrary choices of (hyper)parameters sufficiently justified?" Theoretical beauty would be nice, but that's secondary to a robust result.
If I were reviewing this paper, I'd be mostly satisfied with it. My questions above are honest matters of curiosity rather than strict demands for greater rigour, and "we haven't looked at this" is an acceptable answer to such a question.
I've been a peer reviewer before, albeit not in the machine learning space. My comments above were general questions and not a review, since I'm now curious about applying this technique in my own work.
When I do review a paper, my guide star is "does the paper answer its own question using a methodology powerful enough to detect if the answer is 'no'?" A secondary question is "are any arbitrary choices of (hyper)parameters sufficiently justified?" Theoretical beauty would be nice, but that's secondary to a robust result.
If I were reviewing this paper, I'd be mostly satisfied with it. My questions above are honest matters of curiosity rather than strict demands for greater rigour, and "we haven't looked at this" is an acceptable answer to such a question.