TensorLayer (TuneLayer) 实现 DropConnect

DropConnect 是 Hinton DropOut 之后出现的一种 Regularization 方法，相比DropOut的好处是，它在模型较小的情况下依然能保证准确度。苦于网络上很难找到python例子，这里贴出一个用 TensorLayer 实现的代码，以供参考。

Paper of DropConnect

import tensorflow as tf
import tensorlayer as tl # 注意！！！ TensorLayer 现在更名为 TuneLayer 了 https://github.com/zsdonghao/tunelayer
from tensorlayer.layers import set_keep
import numpy as np
import time

X_train, y_train, X_val, y_val, X_test, y_test = \
                                    tl.files.load_mnist_dataset(shape=(-1,784))

sess = tf.