Errata in print edition
Preface
- p.xvii: “prerequesite” -> “prerequisite”
Chapter 2
- p.12: X-value should be X-value
- p.17: “see why this is rule is” should be “see why this rule is”
- p.17: The expression for \hat Y in Lemma 1 requires \mathit{loss}(0,1)>\mathit{loss}(1,1).
- p.26: The grey dash-dot ROC curve should connect (0,0) to (0.3,1.0) instead of (0.2,1.0). Otherwise the curves will not have equal area.
Chapter 3
- p.38: The equality should be \langle w_{t+1}, x_i\rangle = \langle w_t, x_i\rangle + \|x_i\|^2\,. The printed version has a y_i in front of the \|x_i\|^2 term. This is incorrect.
Chapter 4
- p.56: in the sentence that ends “…the goal in feature design will be to find lifts into high dimensional space such that our data matrix X has linearly independent columns.” columns should be rows.
Chapter 5
p.83 The final equation on the page defining the momentum update should read w_{k+1}=w_{k}-\alpha_{k}g_k\left(w_{k}\right) + \beta (w_{k}-w_{k-1})\,, The printed version has a plus sign in the momentum expression instead of a minus sign.
p.95: The final equation on the page should be (X^T X+ \lambda I_d)^{-1} X^T = X^T (X X^T+ \lambda I_n)^{-1}\,. The printed version has X rather than X^T immediately following the equals sign.
Chapter 6
- p.106: Notation \hat\mu_Z^{(n)} not defined. This is the sample mean of n i.i.d. samples drawn from Z.
- p.116: We claim that all hyperplanes of small norm had low VC dimension. But the hyperplanes have to satisfy y w^T x \geq 1.
Chapter 7
- p.126: “structure linear maps” should be “structured linear maps”
- p.129: “If \epsilon_t=0 is sufficiently small” should be “If \epsilon_t is sufficiently small”
- p.130: “Jacobian matrix is the n\times m matrix” should be “m\times n matrix”
- p.131: \mathsf{D}_{z_{L}} z_L(v_L)= I should be \mathsf{D}_{z^{(L)}} z^{(L)}(v_L)
- p.135: P^TE Q^T should be PEQ
- p.135: E is not defined. It should be the error matrix E=A-B.
Chapter 9
- p.179: Some samples shown in the table are not consistent with the data-generating procedure. The rows 1 1 1 and 1 1 0 cannot occur.
Chapter 11
- p.221: \mathrm{Tr}(Q\Sigma) should be \mathrm{Tr}(M\Sigma)
Chapter 12
- p.251: V_\star(x) \in [0, (1-\gamma)^{-1}] should be V_\star(x) \in [0, 1]
- p.252: \gamma^k should be \gamma^{i-1}
- p.257: “eigenvalue greater than 0 if t > 0.02” should be “eigenvalue greater than 1 if t > 0.02”
Chapter 14
- p.266: there should be no double quote after “pd”
- p.266: “a eigenvalue” should be “an eigenvalue”
- p.274: \lambda_s should be in the denominator
Bibliography
- Reference 131: “continous” should be ”continuous”
Last updated: Mon Oct 27 14:55:25 EDT 2025