ramon-astudillo
commented
9 years ago Yes, please correct those two. On May 26, 2015 7:03 AM, "davidbp" notifications@github.com wrote:
Page 101 form the draft_may25.pf
[image: screen shot 2015-05-26 at 12 46 42] https://cloud.githubusercontent.com/assets/4223580/7810771/4bc56ed2-03a5-11e5-95f3-295999140f03.png
- figure 7.1 states that z = l(x)
- (7.2) states that z = g(x).
Should we change figure 7.1 and write z = g(x) ?
Page 101 (this is not a typo but an extrage way to write the bias). Why don't we write the bias as b instead of w? In the logistic regression example in theano's website the bias is written as b.
— Reply to this email directly or view it on GitHub https://github.com/LxMLS/lxmls_guide/issues/62.
1) Page 101 form the draft_may25.pf
2) Should we change figure 7.1 and write z = g(x) ?
3) sum over k' from 1 to K ( 1=k' must be changed by k' =1 in the underscript of the sum): ![screen shot 2015-05-26 at 17 05 11]
4) Equation 7.4 p(y_k |x ) =( f_k o g_k) (x) instead of p(y_k |x ) =( f_k o g) (x) since g (x) is the net input (vector) and f_k is applied to the k'th component of the net input. Therefore f_k( g_k (x) ) = p(y_k | x)
(https://cloud.githubusercontent.com/assets/4223580/7815554/6701beb6-03c9-11e5-9a8b-d1c9e2c64b54.png)
Maybe this should be rewritten as a vector to enfasize and faciltate reading. For example:
note) Page 101 (this is not a typo but an extrage way to write the bias). Why don't we write the bias as b instead of w? In the logistic regression example in theano's website the bias is written as b.