>The LSTM architecture does permit a correct, 'general' solution to exist, and the authors show that the general solution is an optimum when using a minimum-des

>The LSTM architecture does permit a correct, 'general' solution to exist, and the authors show that the general solution is an optimum when using a minimum-description-length error function. The MDL error function is effectively entropy(weights) + relative entropy(training data | model). >The researcher's problem is that entropy(weights) is not well-defined. I never thought I would say this, but it sounds like a job for Bayesian neural nets. The prior or posterior entropy of the weights should be well-defined in that setting.