AChcxchCA的博客

batchnorm_forward:相关的计算求导，参阅：（ step4累加单元的反向传播划个重点 ）https://kratzert.github.io/2016/02/12/understanding-the-gradient-flow-through-the-batch-normalization-layer.htmldef batchnorm_forward(x, gamma, beta,...

max_pool_forward_naive:def max_pool_forward_naive(x, pool_param): """ A naive implementation of the forward pass for a max pooling layer. Inputs: - x: Input data, of shape (N, C, H, W...

conv_forward_naive:def conv_forward_naive(x, w, b, conv_param): """ A naive implementation of the forward pass for a convolutional layer. The input consists of N data points, each with C ...

spatial_batchnorm_forward:def spatial_batchnorm_forward(x, gamma, beta, bn_param): """ Computes the forward pass for spatial batch normalization. Inputs: - x: Input data of shape (N, ...

import torchimport torch.nn as nnfrom torch.autograd import Variableimport torchvisionimport torchvision.transforms as Timport PILimport numpy as npfrom scipy.misc import imreadfrom collect...

cs231n的官方课程笔记里，有提到如下内容：（译文来源：知乎https://zhuanlan.zhihu.com/p/22038289?refer=intelligentunit）不使用汇聚层：很多人不喜欢汇聚操作，认为可以不使用它。比如在Striving for Simplicity: The All Convolutional Net一文中，提出使用一种只有重复的卷积层组成的结构，抛弃汇...

保存一段可视化滤波器/卷积核的代码：from cs231n.vis_utils import visualize_gridgrid = visualize_grid(model.params['W1'].transpose(0, 2, 3, 1))plt.imshow(grid.astype('uint8'))plt.axis('off')plt.gcf().set_size_inche...

这是一个全连接网络，结构为：{affine - [batch norm] - relu - [dropout]} x (L - 1) - affine - softmax（也是fc_net.py里面的一个网络，纯自己写）class FullyConnectedNet(object): """ Author::Chenx """ """ A fully-con...

这是一个简单的两层网络：affine - relu - affine - softmax（关于网络结构的具体描述都在函数文档里了）from builtins import rangefrom builtins import objectimport numpy as npfrom cs231n.layers import *from cs231n.layer_utils import *...

关于梯度检测的一些代码及使用示例：五种gradient_check:from __future__ import print_functionfrom builtins import rangeimport numpy as npfrom random import randrangedef eval_numerical_gradient(f, x, verbose=True, h=0...

softmax_cross_entropy_loss:（交叉熵损失）def softmax_loss(x, y): """ Computes the loss and gradient for softmax classification. Inputs: - x: Input data, of shape (N, C) where x[i, j] is the ...

svm_hinge_loss: (折叶损失）def svm_loss(x, y): """ Computes the loss and gradient using for multiclass SVM classification. Inputs: - x: Input data, of shape (N, C) where x[i, j] is the sco...

觉得 solver.py 拿来训练很好用，存一波：目录：solver.py  ，  solver训练FullyConnectedNet  ，  外加一段可视化loss与accuracy函数的代码  。solver.py:from __future__ import print_function, divisionfrom builtins import rangefrom builtins ...

dropout_forward:def dropout_forward(x, dropout_param): """ Performs the forward pass for (inverted) dropout. Inputs: - x: Input data, of any shape - dropout_param: A dictionary wi...

relu_forward:def relu_forward(x): """ Computes the forward pass for a layer of rectified linear units (ReLUs). Input: - x: Inputs, of any shape Returns a tuple of: - out: Out...

affine_forward:def affine_forward(x, w, b): """ Computes the forward pass for an affine (fully-connected) layer. The input x has shape (N, d_1, ..., d_k) and contains a minibatch of N ...

课程笔记及其翻译的网址：https://zhuanlan.zhihu.com/p/21930884?refer=intelligentunit一、SGD：def sgd(w, dw, config=None): """ Performs vanilla stochastic gradient descent. config format: - learning_r...