How to stabilize the output of an LSTM?
3 views (last 30 days)
I'm working on a LSTM model which has 4 predictors and 3 classes. At every run, my output predictions are varying. To the best of my knowledge, it is happening so due to the input weights and bias values which are randomized at every run hence leading to unstable predictions. I want to know how to disable this random selection and obtain constant predictions?
Also, I'm using MATLAB 2018a version. There are some training and layer parameters that are present in 2019 and not in 2018.
Any suggestion would be helpful. Thank you.