An python implementation of tiny RNN without framework.
This tiny RNN tackles a simple classification task, by outputing "True" if the prefix sum of the sequential input is greater than a certain value.
This python implementation uses the following formula annotation for readability:
The main functional code is in cnn.py.
Use python3 to run it. "python3 cnn.py"
The accuracy on testing data is 0.097500
The accuracy on testing data is 0.153333
The accuracy on testing data is 0.320000
The accuracy on testing data is 0.886667
The accuracy on testing data is 0.900833
The accuracy on testing data is 0.903333
...