Unitary neural network is able to solve gradient vanishing and gradient explosion problem and help learning long term dependency. EUNN is an efficient and strictly enforced unitary parametrization based on SU(2) group. This repository contains the implementation of Efficient Unitary Neural Network(EUNN) in tensorflow.
If you find this work useful, please cite arXiv:1612.05231.
Don't hesitate to email me if you have any questions :)
requires TensorFlow > 1.2.0
./demo.sh
To use EUNN in your model, simply copy EUNN.py.
Then you can use EUNN in the same way you use built-in LSTM:
from EUNN import EUNNCell
cell = EUNNCell(n_hidden, capacity=2, fft=False, complex=True)
Args:
n_hidden
:Integer
.capacity
:Optional
.Integer
. Only works for tunable style.fft
:Optional
.Bool
. IfTrue
, EUNN is set to FFT style. Default isFalse
.complex
:Optional
.Bool
. IfTrue
, EUNN is set to complex domain. Default isTrue
.
Note:
- For complex domain, the data type should be
tf.complex64
- For real domain, the data type should be
tf.float32
Two tasks for RNN in the paper are shown here. Use -h
for more information
python copying_task.py --model EUNN -T 200
python mnist_task.py --model EUNN -I 20000 -H 512 -C False