Using LSTM to predict Random Numbers from Java's Random() class

Hi, I was interested to try using LSTM to predict Random numbers from Java’s Random class. I use random.getNextDouble() which gives you a random number between 0 and 1. I put these results in one sequence file for the features, and the next number from the sequence in a target file. I trained the network on this data and then evaluated with regression. What I find is the Rsquared is low etc. Am I doing something wrong here? Is it even possible to do this with random data?

This is the code: (which is just one of the examples updated to work)
This is the code for the random generator: