coming from the machine learning research community, i am in awe of the availability and relative ease-of-use of the deep learning frameworks. Rarely can i find

coming from the machine learning research community, i am in awe of the availability and relative ease-of-use of the deep learning frameworks. Rarely can i find comparison code in ML that i didn't have to bug the author for, or try and implement myself based on the paper alone. in short DL is on a much better path than the author realizes. Perhaps we have to thank the github/bitbucket era. The real problem with DL is that until there is a more robust theory(if even possible, the bane and boon of ML is the complexity of the models), much of the application research will sorta be a form of digital alchemy.