# Brml: Errata

The details of the corrections are given in the online version, where you can search for magenta text and a @@' or ++' in the margin. The page numbers refer to the CUP published version.

• pg 11, example 1.5. Updated this -- see the online version.
• pg 12, example 1.7: added a clarifying sentence -- see online version.
• pg 21, exercise 1.18. I should have added that in the absence of any knowledge about the probabilities, Sally should assume maximal uncertainty.
• pg 23, definition 2.5. Should be parents of its children (excluding itself)'
• pg 27 and elsewhere simplical' should be replaced with simplicial'.
• pg 30, paragraph before equation 3.1.1. Changed the probability value -- see online version.
• pg 36, beginning section 3.2.2 updated. See online version.
• pg 40, figure 3.6. (c) with x1->x2
• pg 60, example 4.1, BM-> Boltzmann machine
• pg 61, line before equation 4.2.6, and after 4.2.8, modified according to online version.
• pg 60, figure 4.5. Onsager'.
• pg 65, procedure 4.2. Printing error. The first Ancestral Graph' description has been included in error by the typesetters.
• pg 67, equation 4.2.30. Printing error. There is a missing ')' in the second factor in brackets [..].
• pg 68, equation 4.3.5. The definition needs to be modified to include directed terms. Also additional clarification in the text -- see online version.
• pg 69. Equations 4.3.7, 4.3.8, 4.3.9 -- it's more clear to write p(c|a) instead of \phi(a,c) and similarly p(d|b) instead of \phi(d,b)
• pg 70, after equation 4.4.2, added a clarification sentence.
• pg 70, figure 4.8. Use italics for all node labels.
• pg 72,73 example 4.4. Equation 4.5.9 is missing a term and equation 4.5.10 is incorrectly stated. Please see online version for the correction.
• pg 105, near top: deleted a sentence.
• pg 109, this is still a clique tree (forest);'
• pg 114, figure 6.8 caption. For part (d) this should read We eliminate g and h since these are simplical.'
• pg 115, Procedure 6.2 should read '(Junction Tree algorithm)' (printing error).
• pg 122, equation 6.8.3. Need to remove the trailing ',' in 'a,b,c,'
• pg 134, Causal Consistency paragraph. Deleted the last two sentences.
• pg 137, towards bottom: added '(inverse of the)' -- see online version.
• pg 122. Clarified in definition 6.11 that the junction tree must be consistent.
• pg 136, the ID then corresponds to'. Also deleted a sentence in the Causal Consistency section. See online version.
• pg 137, equation 7.4.11. There is a missing term -- see the online version for the correction.
• pg 139, arrows missing in figure for example 7.6.
• pg 140, equation 7.4.21. This should read =p(x_3|x_2,d_2)', not '=1'.
• pg 141, figure 7.8. Node x_4 should not be shaded green.
• pg 145, example 7.7. Added clarification regarding the figure references -- see online version.
• pg 154, 155 The section From model-based to model-free learning' contains some errors. See the online version for the corrections.
• pg 159, exercise 7.4. I've added some clarifications about moves that would otherwise take the plane out of the airspace.
• pg 168, definition 8.9. Should be everywhere except'.
• pg 189, printing error section 8.9, rule rule' -> rule'
• pg 211, printing error in paragraph before example 9.3. the marginal table equation (9.3.20)' should be the marginal table equation (9.4.19)'.
• pg 212, equation 9.4.26. This last expressions should be (1+1)/(2+2)=1/2
• pg 216 section 9.5.1, 2nd paragraph. Should read limits the number of independence'
• pg 218. The equation after eq 9.5.1 is incorrect -- see the online version. This equation is also missing a number.
• pg 240. Question 9.1 part 3. I should add that the Bayesian tables should be used for the remainder of the question.
• pg 267, algorithm 11.2, line 9. The normalisation 1/N is incorrect. It should be as in equation 11.2.43.
• pg 272. The derivation got a bit garbled here. Please see the online correction, equation 11.4.1.
• pg 281, exercise 11.2. This should be Derive the EM update for the table p(x1|x2,x4).'
• pg 289, equation 12.4.7. The \pi expressions in the final two terms can be removed. The first term should have a minus sign in front to give -N\log(2\pi\sigma^2)
• pg 291 paragraph before section 12.5.2. Should be there will be cases'.
• pg 300, equations 12.9.6 and 12.9.7. The first normalisation term in each expression should be (2\pi\alpha^{-1})^{-K/2}
• pg 332, 2nd line after eq 15.2.15: remove and'.
• pg 338, eq 15.3.7. The 2' superscript in the last term should not be there.
• pg 342, eq 15.5.2. This is is incorrect -- see the online version and corrected sentence before.
• pg 346, after eq 15.6.5. See online version for correction.
• pg 378, eq 17.4.17. The term p(x^n) should not be present in this equation and the preceding (which is also missing a number).
• pg 385, after equation 17.5.11 should be which is to be minimised with respect to boldface(w), ...'
• pg 395, eq 18.1.18. There's a missing summation and $\pi$'s are inverted. See the online version for the correction.
• pg 395, eq 18.1.19. Sign error. Should be +log(det(S)).
• pg 450, after eq 20.5.6, this should be that if $\pi_k$ is low'
• pg 454, section 20.6.3, 2nd paragraph. Should be the lower nodes represent documents, and the upper nodes words'.
• pg 503, example 23.4. Printing error. Remove (robot localisation)'.
• pg 509, equation 23.4.4. The exponent is wrong. See the online version for the correction.
• pg 517, after equation 23.8.19. This should be and that the result of the summation'
• pg 517, equation 23.8.24. The summation over h(t+1) should not be there.
• pg 540, 3rd sentence in section 24.6, this should be s=1,...,S (not s=1,...,d)
• pg 577, sentence before section 26.4.1 should be not limited to the ...'
• pg 596, equation 27.3.7. This should be \sum_i q(i) \int_x q(x'|x,i)p(x) on the right.
• pg 612, eq 27.6.29 should have a normalisation factor. Also the paragraph afterwards contains an incorrect statement; the second and third sentences should be deleted.
• pg 620, after eq 28.3.14. Should be will be of order 1'.
• pg 620, after eq 28.3.15, should be ... and $s^2$ we obtain:'
• pg 624, eq 28.4.15. The sum should be over all j not equal to i.
• pg 638. Printing error. After validity of loopy belief propagation'. Should be For a Markov network..'
• pg 667, section A.5.1. There are some mathematical errors in the section beginning `An unfortunate aspect...' See the online version for the corrections.
• pg 671, section A.5.7. There are some mathematical errors in equation A.5.27 and the following few lines. See the online version for the corrections.