Performance of Deep and Shallow Neural Networks, the Universal Approximation Theorem, Activity Cliffs, and QSAR Authors

[swamidass]

http://onlinelibrary.wiley.com/doi/10.1002/minf.201600118/full

[swamidass]

"Clearly, for the large and diverse drug data sets modelled here using DNN and shallow regularized neural networks (BNN), the predictive power of the models is similar. "

@agitter see what I mean??

[agitter]

Very interesting. I barely skimmed it for now, but it looks it will be quite relevant.