Hi, I have a question about sentences of the page 102 of chapter 5.1.3. Regarding the formula 5.1, it says: this decomposition means that we can solve the ostensibly unsupervised problem of mo...

Definition of the expectation from (3.9) assumes summation across all possible values of x (or taking the integral for all possible x values in the case of a continuous variable). It's only po...

Okay, scaling (5.58) by a factor 1/m gives argmax{1/m*sum(log(p_model(x;theta)))}, that's clear. What still unclear is how the last expression turns into argmax{E_x~p^hat_data(log(p_model(x;th...

I am confused by equation 3.28 in section 3.9.5, used to define an empirical distribution. According to the text, when dealing with discrete variables, p(x) is simply the frequency of that val...

Hello ,all.

I have a question with the "3.9 mixture model", I have no idea to understand the caption of figure 3.2 .

http://www.deeplearningbook.org/contents/prob.html#pf10

Section 3.8: "Expectation, Variance and Covariance" the example in the final lines in the section

I have three questions:

Since shape(D) is (n,l), and transpose of D is (l,n), after left multiply D with its transpose, the output shape will be(l,l), and the constraint of all columns of D are orthonormal is essential if you wanna simplify the matrix myltiplication

Because the decoding function is not necessarily linear. Check kernel PCA.

Can anyone tell me that why the authors say that "PCA is defined by our choice of the decoding functions." in section 2.12 ? Thanks any reply from anyone.

Here's a bit more detailed proof (the moderator can check for any errors in the proof):

consider this:

hi everyone, just joined the community. Excited to be here and learn more about this field.

Is it only me but wouldn't a quadratic function better fit the data points than the linear regression line?

I've always wondered about deep learning's history, and had read in several articles that neural networks and back-propagation used to be popular in 1980s. Didn't know that there was another deep learning phase in the 1960s!

In softmax equation 4.1, the book says:

Consider what happens when all the xi are equal to some constantc. Analytically,we can see that all the outputs should be...