I am experimenting with the TensorFlow dropout function. Since a functionality where the dropout rate decreases with time during training was too complicated to implement (tried all day yesterday), I thought using random dropout rates for each iteration could also be a good idea.
I tried the following line:
X = tf.nn.dropout(X, tf.Variable(tf.random_uniform([], 0.4,0.95)))
For using random dropout rates between 0.4 and 0.95 during training. This did not work out, I got the following error:
FailedPreconditionError: Attempting to use uninitialized value generator_1/Variable
How can I fix this?