先記錄一個很好用的畫神經網絡圖的網站:http://alexlenail.me/NN-SVG/index.html
然後因爲對神經網絡的幾個層的名字到底應該標註在哪有點疑惑,現在看了幾段代碼才弄清楚,所以標註在圖上記錄一下,如下圖(激活函數以ReLU爲例),如果錯誤歡迎指正
上圖中的神經網絡可叫做雙層(應該是雙全連接層)神經網絡或者單隱藏層(one hidden layer)網絡。網絡的前向傳播方式爲輸入節點與W1權重矩陣相乘並加上偏置b1,得到隱藏層的輸入值,然後隱藏層的輸入值需要經過ReLU函數處理,得到隱藏層的輸出函數,然後再用輸出函數重複上述過程,乘以W2權重矩陣並加上偏置b2,得到輸出層的值。在這個過程中需要注意矩陣的維度方向,很容易顛倒出錯,導致維度不一致無法相加或者相乘。
1. neural_net.py
1.1 Q1
第一個問題,在第一次求Loss的地方,出現了這個結果
我的代碼每次跑都是0.018,感覺這個差值有點大了,然後去網上看別人的代碼都是e-13級別的差值,然後在代碼裏找問題找了好久,實在找不出來錯誤,然後用別人的代碼跑也是上面0.018這個結果,最後在這篇博客https://blog.csdn.net/kammyisthebest/article/details/80377613中看到,人家的reg都是0.1,我們的是0.05???然後reg改成0.1跑了一遍,果不其然
1.2 Q2
第二個問題,在求解神經網絡反向傳播的梯度代碼中,遇到一個問題,求W1/W2的梯度都不難理解,但是求b1/b2的梯度時候就遇到問題了,首先代碼中前向傳播是這樣寫的
# 輸入值與W1的點積,作爲下一層的輸入
z2 = X.dot(W1) + b1
# 激活函數,求得隱藏層的輸出,也就是ReLU
a2 = np.maximum(z2, 0)
# 隱藏層的輸出進入到輸出層
scores = a2.dot(W2) + b2
這樣乍得一看,好像b1/b2的偏導數都是數字1,這導致我第一次寫b1/b2的時候直接把偏導寫成了np.ones_like(b1/b2),後來想想,不對啊,這只是在代碼中用numpy庫的簡化寫法罷了,實質上應該這麼寫
# 其實偏置本來也應該是一個矩陣,但是在Numpy的計算中直接被簡化了
z2 = X.dot(W1) + np.ones(N).dot(b1.reshape(H, -1)) # 只是表達這個意思,代碼不一定能跑
a2 = np.maximum(z2, 0)
scores = a2.dot(W2) + np.ones(N).dot(b2.reshape(C, -1))
也就是說在求b1/b2的偏導數時候,實質上求導應該得到的是一個np.ones(N)這麼一個向量,然後再根據cs231n中的求導法則,用上游傳回來的偏導值乘以本地函數值,就可以得到梯度,也就是下面的代碼
# 先求出輸出層softmax型的loss func對輸出層的偏導數,作爲反向傳播的起點,此處與SVM相同
# softmax公式爲L=-s[yi]+ln(∑e^s[j]),可以求得L對s[yi]的偏導數爲-1+e^s[yi]/∑e^s[j],也就是下面代碼中的-1+prob
# 由於輸出層的z和a是相同的值(即a==z),所以此處delta(L)/delta(a) == delta(L)/delta(z)
output = np.zeros_like(scores)
output[range(N), y] = -1
output += prob
# 先根據反向傳播的上層梯度乘以本地變量求出W2的梯度
grads['W2'] = (a2.T).dot(output) # 公式BP4
grads['W2'] = grads['W2'] / N + reg * W2
# 求取b2的梯度,方法同上
grads['b2'] = np.ones(N).dot(output) / N
1.3 代碼
from __future__ import print_function
from builtins import range
from builtins import object
import numpy as np
import matplotlib.pyplot as plt
from past.builtins import xrange
class TwoLayerNet(object):
"""
A two-layer fully-connected neural network. The net has an input dimension of
N, a hidden layer dimension of H, and performs classification over C classes.
We train the network with a softmax loss function and L2 regularization on the
weight matrices. The network uses a ReLU nonlinearity after the first fully
connected layer.
In other words, the network has the following architecture:
input - fully connected layer - ReLU - fully connected layer - softmax
The outputs of the second fully-connected layer are the scores for each class.
"""
def __init__(self, input_size, hidden_size, output_size, std=1e-4):
"""
Initialize the model. Weights are initialized to small random values and
biases are initialized to zero. Weights and biases are stored in the
variable self.params, which is a dictionary with the following keys:
W1: First layer weights; has shape (D, H)
b1: First layer biases; has shape (H,)
W2: Second layer weights; has shape (H, C)
b2: Second layer biases; has shape (C,)
Inputs:
- input_size: The dimension D of the input data.
- hidden_size: The number of neurons H in the hidden layer.
- output_size: The number of classes C.
"""
self.params = {}
self.params['W1'] = std * np.random.randn(input_size, hidden_size)
self.params['b1'] = np.zeros(hidden_size)
self.params['W2'] = std * np.random.randn(hidden_size, output_size)
self.params['b2'] = np.zeros(output_size)
def loss(self, X, y=None, reg=0.0):
"""
Compute the loss and gradients for a two layer fully connected neural
network.
Inputs:
- X: Input data of shape (N, D). Each X[i] is a training sample.
- y: Vector of training labels. y[i] is the label for X[i], and each y[i] is
an integer in the range 0 <= y[i] < C. This parameter is optional; if it
is not passed then we only return scores, and if it is passed then we
instead return the loss and gradients.
- reg: Regularization strength.
Returns:
If y is None, return a matrix scores of shape (N, C) where scores[i, c] is
the score for class c on input X[i].
If y is not None, instead return a tuple of:
- loss: Loss (data loss and regularization loss) for this batch of training
samples.
- grads: Dictionary mapping parameter names to gradients of those parameters
with respect to the loss function; has the same keys as self.params.
"""
# Unpack variables from the params dictionary
W1, b1 = self.params['W1'], self.params['b1']
W2, b2 = self.params['W2'], self.params['b2']
N, D = X.shape
# Compute the forward pass
scores = None
#############################################################################
# TODO: Perform the forward pass, computing the class scores for the input. #
# Store the result in the scores variable, which should be an array of #
# shape (N, C). #
#############################################################################
# *****START OF YOUR CODE (DO NOT DELETE/MODIFY THIS LINE)*****
# 輸入值與W1的點積,作爲下一層的輸入
z2 = X.dot(W1) + b1
# 激活函數,求得隱藏層的輸出,也就是ReLU
a2 = np.maximum(z2, 0)
# 隱藏層的輸出進入到輸出層
scores = a2.dot(W2) + b2
pass
# *****END OF YOUR CODE (DO NOT DELETE/MODIFY THIS LINE)*****
# If the targets are not given then jump out, we're done
if y is None:
return scores
# Compute the loss
loss = None
#############################################################################
# TODO: Finish the forward pass, and compute the loss. This should include #
# both the data loss and L2 regularization for W1 and W2. Store the result #
# in the variable loss, which should be a scalar. Use the Softmax #
# classifier loss. #
#############################################################################
# *****START OF YOUR CODE (DO NOT DELETE/MODIFY THIS LINE)*****
# 根據softmax的Loss函數定義來求該網絡的Loss
# 先減去最大值防止數值錯誤
scores -= np.max(scores, axis=1, keepdims=True)
# 求所有得分項求自然指數
exp_scores = np.exp(scores)
# 求概率矩陣
prob = exp_scores / np.sum(exp_scores, axis = 1, keepdims=True)
# 取出分類正確項的概率
correct_items = prob[range(N), y]
# 根據softmax的loss func求loss
data_loss = -np.sum(np.log(correct_items)) / N
reg_loss = 0.5 * reg * (np.sum(W1 * W1) + np.sum(W2 * W2))
loss = data_loss + reg_loss
pass
# *****END OF YOUR CODE (DO NOT DELETE/MODIFY THIS LINE)*****
# Backward pass: compute gradients
grads = {}
#############################################################################
# TODO: Compute the backward pass, computing the derivatives of the weights #
# and biases. Store the results in the grads dictionary. For example, #
# grads['W1'] should store the gradient on W1, and be a matrix of same size #
#############################################################################
# *****START OF YOUR CODE (DO NOT DELETE/MODIFY THIS LINE)*****
# 先求出輸出層softmax型的loss func對輸出層的偏導數,作爲反向傳播的起點,此處與SVM相同
# softmax公式爲L=-s[yi]+ln(∑e^s[j]),可以求得L對s[yi]的偏導數爲-1+e^s[yi]/∑e^s[j],也就是下面代碼中的-1+prob
# 由於輸出層的z和a是相同的值(即a==z),所以此處delta(L)/delta(a) == delta(L)/delta(z)
output = np.zeros_like(scores)
output[range(N), y] = -1
output += prob
# 先根據反向傳播的上層梯度乘以本地變量求出W2的梯度
grads['W2'] = (a2.T).dot(output) # 公式BP4
grads['W2'] = grads['W2'] / N + reg * W2
# 求取b2的梯度,方法同上
grads['b2'] = np.ones(N).dot(output) / N
# 將最後一層節點的誤差反向傳播至隱藏層
hidden = output.dot(W2.T)
# 考慮到ReLU函數的作用,可以知道只有在z2矩陣中大於零的部分纔會被傳遞至後面的層中,這裏求的就是ReLU函數的偏導矩陣
mask = np.zeros_like(z2)
mask[z2 > 0] = 1
hidden = hidden * mask # N*H,這裏相當於求解出了how bp algorithm works那一章中的公式BP2
# 再從隱藏層反向傳播至W1
grads['W1'] = (X.T).dot(hidden) # 公式BP4
grads['W1'] = grads['W1'] / N + reg * W1
# W1同理
grads['b1'] = np.ones(N).dot(hidden) / N
pass
# *****END OF YOUR CODE (DO NOT DELETE/MODIFY THIS LINE)*****
return loss, grads
def train(self, X, y, X_val, y_val,
learning_rate=1e-3, learning_rate_decay=0.95,
reg=5e-6, num_iters=100,
batch_size=200, verbose=False):
"""
Train this neural network using stochastic gradient descent.
Inputs:
- X: A numpy array of shape (N, D) giving training data.
- y: A numpy array f shape (N,) giving training labels; y[i] = c means that
X[i] has label c, where 0 <= c < C.
- X_val: A numpy array of shape (N_val, D) giving validation data.
- y_val: A numpy array of shape (N_val,) giving validation labels.
- learning_rate: Scalar giving learning rate for optimization.
- learning_rate_decay: Scalar giving factor used to decay the learning rate
after each epoch.
- reg: Scalar giving regularization strength.
- num_iters: Number of steps to take when optimizing.
- batch_size: Number of training examples to use per step.
- verbose: boolean; if true print progress during optimization.
"""
num_train = X.shape[0]
iterations_per_epoch = max(num_train / batch_size, 1)
# Use SGD to optimize the parameters in self.model
loss_history = []
train_acc_history = []
val_acc_history = []
for it in range(num_iters):
X_batch = None
y_batch = None
#########################################################################
# TODO: Create a random minibatch of training data and labels, storing #
# them in X_batch and y_batch respectively. #
#########################################################################
# *****START OF YOUR CODE (DO NOT DELETE/MODIFY THIS LINE)*****
# 加上replace=False時候提示Cannot take a larger sample than population when 'replace=False',即batch_size>num_train時錯誤,故去掉
random_index = np.random.choice(num_train, batch_size)
X_batch = X[random_index, :]
y_batch = y[random_index]
pass
# *****END OF YOUR CODE (DO NOT DELETE/MODIFY THIS LINE)*****
# Compute loss and gradients using the current minibatch
loss, grads = self.loss(X_batch, y=y_batch, reg=reg)
loss_history.append(loss)
#########################################################################
# TODO: Use the gradients in the grads dictionary to update the #
# parameters of the network (stored in the dictionary self.params) #
# using stochastic gradient descent. You'll need to use the gradients #
# stored in the grads dictionary defined above. #
#########################################################################
# *****START OF YOUR CODE (DO NOT DELETE/MODIFY THIS LINE)*****
self.params['W1'] -= grads['W1'] * learning_rate
self.params['W2'] -= grads['W2'] * learning_rate
self.params['b1'] -= grads['b1'] * learning_rate
self.params['b2'] -= grads['b2'] * learning_rate
pass
# *****END OF YOUR CODE (DO NOT DELETE/MODIFY THIS LINE)*****
if verbose and it % 100 == 0:
print('iteration %d / %d: loss %f' % (it, num_iters, loss))
# Every epoch, check train and val accuracy and decay learning rate.
if it % iterations_per_epoch == 0:
# Check accuracy
train_acc = (self.predict(X_batch) == y_batch).mean()
val_acc = (self.predict(X_val) == y_val).mean()
train_acc_history.append(train_acc)
val_acc_history.append(val_acc)
# Decay learning rate
learning_rate *= learning_rate_decay
return {
'loss_history': loss_history,
'train_acc_history': train_acc_history,
'val_acc_history': val_acc_history,
}
def predict(self, X):
"""
Use the trained weights of this two-layer network to predict labels for
data points. For each data point we predict scores for each of the C
classes, and assign each data point to the class with the highest score.
Inputs:
- X: A numpy array of shape (N, D) giving N D-dimensional data points to
classify.
Returns:
- y_pred: A numpy array of shape (N,) giving predicted labels for each of
the elements of X. For all i, y_pred[i] = c means that X[i] is predicted
to have class c, where 0 <= c < C.
"""
y_pred = None
###########################################################################
# TODO: Implement this function; it should be VERY simple! #
###########################################################################
# *****START OF YOUR CODE (DO NOT DELETE/MODIFY THIS LINE)*****
# 前向傳播,求出輸出值
z2 = X.dot(self.params['W1']) + self.params['b1']
a2 = np.maximum(z2, 0)
scores = a2.dot(self.params['W2']) + self.params['b2']
# 求出得分矩陣每一行最大值的索引,代表分類的類別
y_pred = np.argmax(scores, axis=1)
pass
# *****END OF YOUR CODE (DO NOT DELETE/MODIFY THIS LINE)*****
return y_pred
2. two_layer_net.ipynb
best_net = None # store the best model into this
#################################################################################
# TODO: Tune hyperparameters using the validation set. Store your best trained #
# model in best_net. #
# #
# To help debug your network, it may help to use visualizations similar to the #
# ones we used above; these visualizations will have significant qualitative #
# differences from the ones we saw above for the poorly tuned network. #
# #
# Tweaking hyperparameters by hand can be fun, but you might find it useful to #
# write code to sweep through possible combinations of hyperparameters #
# automatically like we did on the previous exercises. #
#################################################################################
# *****START OF YOUR CODE (DO NOT DELETE/MODIFY THIS LINE)*****
best_acc = 0
learning_rate = [1e-4, 5e-4, 1e-3]
regulations = [0.2, 0.25, 0.3, 0.35]
for lr in learning_rate:
for reg in regulations:
stats = net.train(X_train, y_train, X_val, y_val,
num_iters=1500, batch_size=200,
learning_rate=lr, learning_rate_decay=0.95,
reg=reg, verbose=True)
val_acc = (net.predict(X_val) == y_val).mean()
if val_acc > best_acc:
best_acc = val_acc
best_net = net
print('lr = ',lr ,' reg = ',reg, ' acc = ', best_acc)
pass
# *****END OF YOUR CODE (DO NOT DELETE/MODIFY THIS LINE)*****
昨天晚上跑的時候在val集上最高的準確率達到了0.527,但是最後的參數出錯,好像因爲learning_rate設置太大導致nan錯誤,不知道爲什麼0.527的best_net也沒有保存下來,今天再跑,最高的準確率只有0.52了,
然後最終在test_set上的測試結果