11-14-2018 09:05 AM
Rather than a measure, this is a python-tool that generates DAX code to implement a pre-trained neural network (or linear/logistic regression) for prediction.
The keras branch generates dax code for Keras models. Currently supports SimpleRNN and Dense layers, with tanh, sigmoid or relu activation functions. Make sure the name of the dense layers contains substring "dense" and the simple rnn layers contains "rnn". (They do by default, unless you rename the layers).
This is just cool as hell - really clever @alekskr. As probably a small set of people that have written a lot of keras *and* a lot of dax... I enjoyed the hell out of this 🙂