Page 53: In the proof of $\text{trace}(AB)=\text{trace}(BA)$, if $A$ is $d\times N$ and $B$ is $N\times d$,
the display should be
$$\sum_{i=1}^d (AB)_{ii}=\sum_{i=1}^d \sum_{k=1}^N a_{ik}b_{ki}=
\sum_{k=1}^N \sum_{i=1}^d b_{ki}a_{ik}=\sum_{k=1}^N (BA)_{kk}.$$
Page 208: In Archimedes' axiom, "than that" should be "then that".
Page 259: In Exercise 4.5.13, for the converse, assume also $f(x)$ is convex.
Page 396: Weights $m$, $w$ are row vectors and samples $x$, $\tilde x$ are column vectors, so the display should be
$$w=(m,b),\qquad \tilde x=\begin{pmatrix} x\\1\end{pmatrix}.$$
Page 411: The weight matrix $M$ and the bias vector $b$ are
$$M= \begin{pmatrix} m_1\\ m_2\\ m_3\end{pmatrix},\qquad b=\begin{pmatrix} b_1\\ b_2\\ b_3\end{pmatrix},\qquad W=(M,b),$$
so $M$ is $3\times4$, $b$ is a column vector in ${\bf R}^3$, and the augmented weight matrix $W$ is $3\times5$.
Page 457: The last sentence before the code should be "a matrix with a single column".
Page 520: At bottom, the last sentence should be "the hermitian product $P\overline{P'}$ of $P$ and $P'$".
Page 543: The definition of closed should be "A set $K$ is closed if $x_n$ in $K$ and $x_n\to x$ implies $x$ is in $K$".
