Closed agitter closed 7 years ago
This paper studies the Merck Kaggle dataset, as in #54 and #57. The main contribution is a comparison of a neural network initialized with Gaussian weights versus networks initialized with a deep belief network (DBN). They assess different sampling strategies for the DBN. DBN initialization does improve predictive performance over their baseline neural network. However, I don't think this adds anything relevant for the objectives of our review if we already discuss #54 and #57.
http://doi.org/10.1002/jcc.24671
Virtual screening (#45) on the Merck Kaggle dataset