search

bisno / WDRO

UniDT鲁棒优化算法 The implementation on Towards Scalable and Fast Distributionally Robust Optimization for Data-Driven Deep Learning.
MIT License
181 stars 26 forks source link

the weight parameter in toydataset. #10

Open hetolin opened 1 year ago

hetolin commented 1 year ago

W = tf.get_variable("w", shape=[1, 1], initializer=tf.contrib.layers.xavier_initializer())

Does the initializer affect model perfomance? Can I change the xavier_initializer()?

bisno commented 1 year ago

Thanks for your comment. As far as we know, you can also try tf.contrib.keras.initializers.VarianceScaling.