Feedback on "Machine learning: a probabilistic perspective"
The version of the browser you are using is no longer supported. Please upgrade to a supported browser.Dismiss

ABCDEFGHIJKLMNOPQRSTU
1
2
4/6/2012 9:40:11621how toconvert ourlmthang@stanford.edufixed
3
4/6/2012 9:40:42622we quantities can computelmthang@stanford.edufixed
4
4/6/2012 13:19:05824The factor on line 2 should probably be for z_e = 1dwu4@stanford.edufixed
5
4/6/2012 13:45:471test-
6
4/6/2012 15:23:27615Paragraph below 18.11 -> p(z1) = N(mu,sigma) not N(z|mu, sigma)-
7
4/6/2012 15:23:39649"The square nodes, represent measurement failures and resistance failures, represent discrete variables."chrisb33@stanford.edufixed
8
4/6/2012 15:27:30621625boldface x_t is used inconsistently as the input signal and as the robot position (e.g. cov[x_y|y_{1:t}], "the robot's path x_{1:t}")chrisb33@stanford.edu-
9
4/6/2012 15:36:47615Just above 18.2.2 -> change Figure 18.1(a) and (b) to (b) and (c) respectivelyfixed
10
4/6/2012 18:01:49793Minor typo, 4th line: "we we just anneal"laureen@stanford.edufixed
11
4/9/2012 18:41:42619The "Q" in "plus observation noise with variance Q" in the last paragraph looks like it should be "R".mabercr@stanford.edufixed
12
4/10/2012 21:44:23249overwhelmed bu the likelihoodlmthang@stanford.edufixed
13
4/10/2012 21:46:33249inconsistent parameter dimension: D in 8.4.1, d in 8.4.2lmthang@stanford.edufixed
14
4/12/2012 9:34:37632We first consider the online case ... We then consider the online caselmthang@stanford.edufixed
15
4/12/2012 12:08:37633For consistency, eq. (18.44), (18.45), (18.48), we should condition on x_{1:t}.lmthang@stanford.edufixed
16
17
4/17/2012 18:14:11772I made a wrong comment above, but Exp and exp are very confusing. Originally I though x|\lambda is a fraction in the exponential function. It might be better just use f(x; \lambda).lmthang@stanford.edufixed
18
4/17/2012 18:27:57773metodlmthang@stanford.edufixed
19
4/19/2012 0:15:21795Link to SIAM News is brokenlmthang@stanford.edufixed
20
4/19/2012 23:00:15824g_e(x_e, z_e=0) .. and g_e(x_e, z_e=0)lmthang@stanford.edufixed
21
4/21/2012 21:33:41968Candès rendered improperly as CandAIsjkrause@cs.stanford.edufixed
22
23
4/25/2012 10:05:56706\defeq for Eq. 21.3lmthang@stanford.edu
24
4/27/2012 21:12:21355-357citation (Liang and Klein Liang and Klein)lmthang@stanford.edu
25
5/8/2012 12:33:46901mycitelmthang@stanford.edu
26
5/8/2012 12:46:15902b_k = B_{:,k}. should be B_{k, :}? as later we have b_{kv}lmthang@stanford.edu
27
5/8/2012 13:55:16902Eq. (27.3) Cat(y_{il} |...). Should it be y_{ir} ?lmthang@stanford.edu
28
5/8/2012 14:03:24903Eq (27.7) y_{1:R}. should be y_{i, 1:R}.lmthang@stanford.edu
29
5/8/2012 14:29:55904Inconsistent notation _{v,k} vs _{vk}. Also indices in the formula of n_{iv} seems wrong?lmthang@stanford.edu
30
5/8/2012 14:36:53901-905Though the presentation is very precise, many notations like L_i vs R, or y_il vs y_ir, n_iv vs y_ir make it tough for readers at first. Perhaps, a table to compare and contrast notations and equations between models with fixed and variable length sequences could be useful.lmthang@stanford.edu
31
5/8/2012 17:26:45906-LDA section - the notations used are markedly different from those in the original LDA paper, which makes it hard to track down corresponding parts. It seems that the notations used in the book are meant to relate previously discussed models, but I find those in the original paper more intuitive. Perhaps, better subscripts could be used, e.g. d -- document, w -- word, t -- topic, instead of i,k, l. Also, why do we use q for latent variables, whereas z seems to be a standard choice, both in the book itself and the LDA paper?lmthang@stanford.edu
32
5/21/2012 15:55:09877"directl link" near the beginning of 26.5.1.3jkrause@cs.stanford.edu
33
5/21/2012 15:57:06966url at end of "Basu" reference spills over into next columnjkrause@cs.stanford.edu
34
5/21/2012 15:59:57879"However, this approach is very efficient" near beginning of 26.5.2jkrause@cs.stanford.edu
35
5/21/2012 16:08:13879eq 26.41: missing a right parenthesis and equation overflows into equation numberjkrause@cs.stanford.edu
36
5/27/2012 23:17:52834Figure 25.2: plate model diagrams look bigger than necessary.lmthang@stanford.edu
37
5/27/2012 23:19:10835Figure 25.3: "from left to right, we show N=50, N=500 and N=1000". But there are only 2 columns ?lmthang@stanford.edu
38
5/28/2012 0:27:31837DP generalizes this to ... p(\theta \in T_i) = H(T_i). I think it should be G(T_i) instead of H(T_i)lmthang@stanford.edu
39
5/28/2012 0:27:58838this process processlmthang@stanford.edu
40
5/28/2012 1:02:58838draw samples form a GPlmthang@stanford.edu
41
5/28/2012 14:16:17334section 11.2.2: latent variables do not have to any meaninglmthang@stanford.edu
42
5/28/2012 15:19:11907This is importance since in natural language' in paragraph beginning with 'One advantage'jkrause@cs.stanford.edu
43
5/28/2012 15:21:16909"to play the coronet" -> "to play the cornet". A coronet is a crown, not an instrumentjkrause@cs.stanford.edu
44
5/28/2012 15:31:30915"In the bathc version" -> "In the batch version", beginning of 27.3.6jkrause@cs.stanford.edu
45
5/28/2012 15:34:08917Dirichelt -> Dirichlet, first paragraph of 27.4.1jkrause@cs.stanford.edu
46
5/28/2012 15:36:53919"infernece" -> "inference", paragraph that begins with "One can perform"jkrause@cs.stanford.edu
47
5/28/2012 15:40:30930atttributes -> attributes (beginning of 27.5.3)jkrause@cs.stanford.edu
48
5/30/2012 23:43:43965authors' names in bibliography entry begins with 'and' (middle column)jkrause@cs.stanford.edu
49
6/2/2012 21:41:11281\theta \phi(x). Should be \top{\theta} \phi(x)lmthang@stanford.edu
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100