吴恩达第5周答案 Neural Networks: Learning 答案

1。

You are training a three layer neural network and would like to use backpropagation to compute the gradient of the cost function. In the backpropagation algorithm, one of the steps is to update

Δ(2)ij:=Δ(2)ij+δ(3)i(a(2))j

for every  i,j . Which of the following is a correct vectorization of this step?

Δ(2):=Δ(2)+(a(2))Tδ(3)

Δ(2):=Δ(2)+δ(3)(a(2))T

Δ(2):=Δ(2)+(a(3))Tδ(2)

Δ(2):=Δ(2)+δ(3)(a(3))T

1
point
2。

Suppose  Theta1  is a 5x3 matrix, and  Theta2  is a 4x6 matrix. You set  thetaVec=[Theta1(:);Theta2(:)] . Which of the following correctly recovers  Theta2 ?

reshape(thetaVec(16:39),4,6)

reshape(thetaVec(15:38),4,6)

reshape(thetaVec(16:24),4,6)

reshape(thetaVec(15:39),4,6)

reshape(thetaVec(16:39),6,4)

1
point
3。

Let  J(θ)=2θ3+2 . Let  θ=1 , and  ϵ=0.01 . Use the formula  J(θ+ϵ)J(θϵ)2ϵ  to numerically compute an approximation to the derivative at  θ=1 . What value do you get? (When  θ=1 , the true/exact derivati ve is  dJ(θ)dθ=6 .)

5.9998

6

6.0002

8

1
point
4。

Which of the following statements are true? Check all that apply.

Gradient checking is useful if we are using gradient descent as our optimization algorithm. However, it serves little purpose if we are using one of the advanced optimization methods (such as in fminunc).

Using gradient checking can help verify if one's implementation of backpropagation is bug-free.

If our neural network overfits the training set, one reasonable step to take is to increase the regularization parameter  λ .

Using a large value of  λ  cannot hurt the performance of your neural network; the only reason we do not set  λ  to be too large is to avoid numerical problems.

1
point
5。

Which of the following statements are true? Check all that apply.

Suppose that the parameter  Θ(1)  is a square matrix (meaning the number of rows equals the number of columns). If we replace  Θ(1)  with its transpose  (Θ(1))T , then we have not changed the function that the network is computing.

Suppose we have a correct implementation of backpropagation, and are training a neural network using gradient descent. Suppose we plot  J(Θ)  as a function of the number of iterations, and find that it is increasing rather than decreasing. One possible cause of this is that the learning rate  α  is too large.

If we are training a neural network using gradient descent, one reasonable "debugging" step to make sure it is working is to plot  J(Θ)  as a function of the number of iterations, and make sure it is decreasing (or at least non-increasing) after each iteration.

Suppose we are using gradient descent with learning rate  α . For logistic regression and linear regression,  J(θ)  was a convex optimization problem and thus we did not want to choose a learning rate  α  that is too large. For a neural network however,  J(Θ)  may not be convex, and thus choosing a very large value of  α  can only speed up convergence.

评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值