LSTM, how to get a probability distribution instead of a predicted value?

Question

I am using a LSTM network for bitcoin forecasts. But instead of predicting a fixed value (i.e, if one use a LSTM to predict bitcoin generally we get a prediction for the price a