Week 1 Quiz - Practical aspects of deep learning（第一周测验 - 深度学习的实践）

\1. If you have 10,000,000 examples, how would you split the train/dev/test set? (如果你有 10,000,000 个样本，你会如何划分训练/开发/测试集？)

【 】98% train . 1% dev . 1% test(训练集占 98% ， 开发集占 1% ， 测试集占 1%)

答案

对

 

\2. The dev and test set should: (开发和测试集应该)

【 】Come from the same distribution (来自同一分布)

答案

对

 

\3.If your Neural Network model seems to have high variance, what of the following would be promising things to try? (如果你的神经网络模型似乎有很高的方差，下列哪个尝试是可能解 决问题的？)

【 】Add regularization(添加正则化)

【 】Get more training data (获取更多的训练数据)

答案

全对

 

\4. You are working on an automated check-out kiosk for a supermarket, and are building a classifier for apples, bananas and oranges. Suppose your classifier obtains a training set error of 0.5%, and a dev set error of 7%. Which of the following are promising things to try to improve your classifier? (Check all that apply.) (你在一家超市的自动结帐亭工作，正在为苹果，香蕉和橘子制作分类器。 假设您的分类器在训练集上有 0.5％的错误，以及开发集上有7％的错误。 以下哪项尝试是有希望改善你的分类器的分类效果的？)

【 】Increase the regularization parameter lambda (增加正则化参数 lambda)

【 】Get more training data (获取更多的训练数据)

答案

全对

 

\5. What is weight decay? (什么是权重衰减？)

【 】A regularization technique (such as L2 regularization) that results in gradient descent shrinking the weights on every iteration. (正则化技术（例如 L2 正则化）导致梯度下降在每次迭代时权重收缩。)

答案

对

 

\6. What happens when you increase the regularization hyperparameter lambda? (当你增加正 则化超参数 lambda 时会发生什么？)

【 】Weights are pushed toward becoming smaller (closer to 0) (权重会变得更小（接近 0）)

答案

对

 

\7. With the inverted dropout technique, at test time: (在测试时候使用 dropout)

【 】You do not apply dropout (do not randomly eliminate units) and do not keep the 1/keep_prob factor in the calculations used in training(不要随机消除节点，也不要在训练中使用的计算中保留 1 / keep_prob 因子)

答案

对

 

\8. Increasing the parameter keep_prob from (say) 0.5 to 0.6 will likely cause the following: (Check the two that apply) (将参数 keep_prob 从（比如说）0.5 增加到 0.6 可能会导致以下 情况)

【 】Reducing the regularization effect (正则化作用减弱)

【 】Causing the neural network to end up with a lower training set error(使神经网络在结束时会在训练集上表现好一些。)

答案

全对

 

\9. Which of these techniques are useful for reducing variance (reducing overfitting)? (Check all that apply.) (以下哪些技术可用于减少方差（减少过拟合）)

【 】Dropout

【 】L2 regularization (L2 正则化)

【 】Data augmentation(数据增强)

答案

全对

 

\10. Why do we normalize the inputs x? (为什么我们要归一化输入 x？)

【 】It makes the cost function faster to optimize(它使成本函数更快地进行优化)

答案

对

 

 

---

---

Week 1 Code Assignments：

✧Course 2 - 改善深层神经网络 - 第一周测验 - 深度学习的实践

✦assignment1_1：Initialization)

https://github.com/phoenixash520/CS230-Code-assignments

✦assignment1_2：Regularization

https://github.com/phoenixash520/CS230-Code-assignments

✦assignment1_3：Gradient Checking

https://github.com/phoenixash520/CS230-Code-assignments