Last update 15 Jun 2006. N.B.: The correct fonts have not been reproduced in this errata. If not otherwise mentioned, no changes in the fonts should be made.

p. 35 1st line: "variance" should be "standard deviation"

p. 41 Eq. (2.101): upper index in sum should be infinity (not "n")

p. 104 Problems 4.17.2 and 4.17.3:
text "given *x*" should be added at the end of the problem statements.

p. 104 Problems 4.18.1 - 4.18.3: the questions should be asking about estimates of "z" and not of "y".

p. 104 Computer assignment 4.1: In the web address "/ mlearn/" should be "/~mlearn/"

p. 108 Eq. (5.5) It should be mentioned for clarity that this a definite integral over the real line. The same applies for any later use of the definition, including the multidimensional case where the integral is taken of the n-dimensional real space.

p. 111 Eq. (5.21) and the next text row: "m" should be "n"

p. 113 Eq. (5.26), first row: "log|M|" should be "log|det M|"

p. 114 Eq. (5.29): on the right-hand side "1" should be "i!".

p. 116 Eq. (5.36): "m" should be "n"

p. 117 Last row before Section 5.6.2: "latter part of (5.39)" should be "(5.40)".

p. 119 Eq. (5.49): "x" should be replaced by Greek "ksi" (twice)

p. 121 Eq. (5.51): there should be a minus sign as the first thing inside the parentheses.

p. 140
2nd row above Eq. (6.33): "C" should be "C_{x}"

p. 157 Fig. 7.7 caption: "if give" should be "is given"

p. 160
Eq. (7.23): "E{ U y y^{T} U^{T} }" should be "E { U z z^{T} U^{T}}"

p. 164 Problem 7.1: Ignore this one, the claim seems to be false (its would require more assumptions and we're not sure which)

p. 164 Computer assignment 7.1, 3rd row: "Plot the distributions of the independent components" should be "Plot the distributions of the observed mixtures".

p. 171 Eq. (8.6): there should be a minus sign as the first thing inside the parentheses.

p. 177
In Eq. (8.13), left hand side, as well as on 2nd line from bottom:
"kurt(w^{T} x)" should be "kurt(w^{T} z)".

p. 183 1st line: "random variable" should be "random vector".

p. 185 Two occurences of "the constant \gamma" should be "the parameter \gamma".

p. 187 Theorem 8.1.: "rows of the mixing matrix" should be "rows of the inverse of the mixing matrix"

p. 189 Eq. (8.43): One closing curly bracket is not in the right place, see Step 4, Table 8.2, p. 190 for the correct place.

p. 193 Fig 8.24: The dashed line that points to the left should be omitted.

p. 196
In both tables, at step 5: the last occurrence of w should have an index: w_{p} and w_{i} , respectively

p. 206
Eq. (9.8): "g'(s_{i})" should be "g'_{i}(s_{i})"

p. 207 line after Eq. (9.13): "g_{i}(y_{i})" should be "g_{1}(y_{1})"

p. 208 3rd line after Eq. (9.17): "**I**" should be "-**I**"

p. 213 line after Eq. (9.33): the "w"'s should be replaced by "b"'s

p. 218 Problem 9.8: "a gaussian random variable" should be "a gaussian random variable of zero mean and unit variance".

p. 230 Eq. (11.1): on the second line, all exponents "4" should be deleted.

p. 233 lines 1 and 2: "matrix", "matrices" should be "vector", "vectors"

p. 233
Eq. (11.10): right-hand side should be F(w w^{T}) w

p. 235 last line: "cumulant" should be "cumulant tensor"

p. 247 in the very middle: "how the nonlinear function g(y) can be determined" should be "how the nonlinear function f can be determined".

p. 262
Ex. 12.1: "f(y_{2}) = y_{2}" should be "g(y_{2}) = y_{2}"

p. 262 Ex. 12.6: in the equation, add B to the end after the closing bracket.

p. 300 Equation (15.18): "0" should be in boldface (it's a vector)

p. 306 Eq. (16.5): there should be a minus sign as the first thing inside the parentheses.

p. 307 Eq. (16.6): "max" should be "min".

p. 308 Line 5: "descent" should be "ascent".

p. 308
Line 8: "using this new A, we compute the ML estimates of the s_{i}(t)"
should be "using these new s_{i}(t), we compute the ML estimate of A".

p. 320 Equation on 5th line: remove last lonely absolute value sign "|"

p. 331 7th line below Eq. (17.31): remove "S," from the in-line equation.

p. 337 1st line: remove the text "(outputs)"

p. 342 Equation (18.4): there should be a "not equal to" sign between i and j

p. 344 2nd line: the matrix should have superscript z.

p. 346 4th row below Eq. (18.20): "Q" should have Greek small tau as subscript.

p. 348 Eq. (18.27): there should be a minus sign just before the summation sigma.

p. 351 2nd line after (18.31): the definition of **C** (with bar on top and some sub and superscripts) is wrong, see (18.7) and (18.2) (replacing x by z) for the correct definition.

p. 353 6th row of subsection on ICA and Kolmogoroff complexity: "MLD" should be "MDL".

p. 359 Equation (19,15) and p. 362 Equation (19.22): the subscript of delta should be "k,0" and the conditions on the right-hand side should be "k=0" and "k not equal to 0".

p. 362 three lines below Equation (19.22): "t=k" should be "k=0"

p. 365 Equation (19.29): subscript of S should be "j".

p. 373 Equation (20.2): there should be a minus sign just before the summation sigma. Same thing two lines below.

p. 375 7th line and p. 377 3rd line: the elements summed should be squared as mentioned in the text.

p. 394 Row 3: "functions" should be "function".

p. 397 Row 7 of 3th full paragraph: "could be interpreted as measures of nongaussianity as well" should be "could interpreted as measures of sparseness as well"

p. 457 reference 147: page numbers should be 99-114