3 ms·
No source code :(
by malux85 9y ago
No source code :(
- callesgg 9y agoRight, they probably just made it all up.
- TehCorwiz 9y agoNo, that's unlikely. But it does mean that independent verification is difficult, if not impossible, given that we don't have their implementation or working data set.
- cycrutchfield 9y agoThe snark is unnecessary. Many papers in deep learning display heavily cherry-picked results. Including a source code implementation is one way to ensure that the real world results match the cherry picked results
- jdc 9y agoHehe
- malux85 9y agoNo, but independent third party verification is a cornerstone of science. I read many many deep learning papers and very often there are cherry picked results OR poor validation methods. For example, it's common for naive/inexperienced people to predict on the same data that the network is trained on : because it gives much better results due to overfitting. I'm not saying that's what this paper has done. But if you are indeed going to publish, then your experiment needs to be reproducible, this means including the source code and the data.
- ppbutt 9y agoRight, nobody would ever lie or stretch the truth right?
- consultutah 9y agoNo code yet, but promised: http://www.cs.cmu.edu/~chenk/ http://www.cs.cmu.edu/~chenk/