qq_45602850的博客

正则化目的：防止过拟合、增强模型的泛化能力。什么是 dropout？在神经网络中，遍历神经网络的每一层，为每一层设置一个概率keep-prob，并以1-(keep-prob)的概率移除一些神经元。 也可以理解为每一个神经元被保留下来的概率为该层的keep-prob。注意：深度学习模型在训练时使用dropout，在测试时不使用dropout。Inverted dropout 反向随机激活随机生成矩阵d[l]=np.random.randn(a[l].shape,a[l-1].shape)判断

Some of the problems below are best addressed using a supervisedlearning algorithm, and the others with an unsupervised learningalgorithm. Which of the following would you apply supervised learningto? (Select all that apply.) In each case,assume some a