-- p. 6, caption of Figure 1.2: the analytical form of the MP density is proportional to $\sqrt{(\tmax(\alpha) - \gamma) (\gamma - \tmin(\alpha))}/\gamma$. I.e., the denominator should be outside the square root, in contrast to the form given in the caption.
-- p. 17: derivation at bottom of page makes use of dominated convergence
in moving the expecation inside the infinite summation from Taylor expansion
-- p. 22, line 4: Should read "exists for all $|\lambda| \leq b$", as opposed to "$\lambda \leq |b|$".
-- p. 23, equation (2.9): should read "for all $t \geq 0$."
-- p. 50: reference should be to p. 78 of Ledoux (2001), following statement of Corollary 4.10. (Page number 128 is incorrect.)
-- p. 52, Exercise 2.10(c): the $\delta$ in the RHS of the inequality in (2.65b) should be replaced by $\tilde\delta$
-- p. 56: Exercise 2.21 (b)(i): should read $\mathbb{P}[V \geq 1] \rightarrow 1$, not tending to infinity.
-- p. 67: Following "after makes use of the fact that...", we should have $e^g$ instead of $g$ within the conditional expectation.
-- p. 140: first line, should be u covering, instead of \delta covering.
-- p. 146: line above equation (5.59), missing a right parenthesis ) around
Y_1, ... Y_n.
-- p. 168, Proof of Theorem 6.5: the bound $(4 \ell)! \leq 2^{2 \ell} [(2 \ell)!]^2$ is incorrect as stated. It should be replaced with the bound $(4 \ell)! \leq [ 2^{2 \ell} \,(2 \ell)!]^2$. With this change, the argument will still go through, but the values of the constants will be changed.
-- p. 215, equation (7.33): the factor $2$ on the last term involving $\|\theta^*_{S^c}\|_1$ should be replaced by a factor $4$. This change will affect the constants in remaining equations of the proof, as well as in the statement of Theorem 7.19.
-- p. 228, "...can achieve the fast rate with only a column normalization condition on the design matrix...". Should read "without any conditions on the design matrix". See Example 13.16 for details.
-- p. 231, Exercise 7.7(b): assumed condition should read as $n \gtrsim s \log (\frac{e d}{s})$. (Flip $s$ and $d$ in the fractional part.)
-- p. 239, equation (8.7): eigenvalues of \Sigma should not be squared. (I.e.., $\gamma^2$ ---> $\gamma$).
-- p. 385: Proof of Riesz representation theorem. "...such that $\|g\|_H = L(g)$..." should be changed to
$\|g\|_H^2 = L(g)$." I.e., add a square!
-- p. 390: in the proof of Theorem 12.11, there is asign error in the polarization identity: the coefficient of \|g\|^2 should be negative, not positive.
-- p. 394, equation (12.10): Last term on right hand side should read $\int_0^1 f^{(\alpha)}(z) (x-z)^{\alpha-1}/(\alpha-1)!$. That is, there should be no $(\cdot)_+$ in this equation; it is simply the integral form of Taylor's theorem.
-- p. 404: proof of Prop. 12.31: needs some discussion of completeness of the constructed tensor product.
-- p. 429: in the derivation for Lipschitz functions, there is a missing
1/sqrt{n} factor in the first entropy integral.
-- p. 455: equation (14.6) is missing the left-side absolute value sign
-- p. 467, Statement of Corollary 14.15: the Rademacher complexity $\delta_n$ inequation (14.32) should be defined in terms of the shifted function class $\mathcal{F}^*$
-- p. 521, Exercise 15.10(b): should read "for any $\gamma \in \mathbb{R}$"