Paper 1, Section II, J

Mathematics of Machine Learning
Part II, 2021

Let H\mathcal{H} be a family of functions h:X{0,1}h: \mathcal{X} \rightarrow\{0,1\} with H2|\mathcal{H}| \geqslant 2. Define the shattering coefficient s(H,n)s(\mathcal{H}, n) and the VCV C dimension VC(H)\mathrm{VC}(\mathcal{H}) of H\mathcal{H}.

Briefly explain why if HH\mathcal{H}^{\prime} \subseteq \mathcal{H} and H2\left|\mathcal{H}^{\prime}\right| \geqslant 2, then VC(H)VC(H)\mathrm{VC}\left(\mathcal{H}^{\prime}\right) \leqslant \mathrm{VC}(\mathcal{H}).

Prove that if F\mathcal{F} is a vector space of functions f:XRf: \mathcal{X} \rightarrow \mathbb{R} with FF\mathcal{F}^{\prime} \subseteq \mathcal{F} and we define

H={1{u:f(u)0}:fF}\mathcal{H}=\left\{\mathbf{1}_{\{u: f(u) \leqslant 0\}}: f \in \mathcal{F}^{\prime}\right\}

then VC(H)dim(F)\operatorname{VC}(\mathcal{H}) \leqslant \operatorname{dim}(\mathcal{F}).

Let A={{x:xc22r2}:cRd,r[0,)}\mathcal{A}=\left\{\left\{x:\|x-c\|_{2}^{2} \leqslant r^{2}\right\}: c \in \mathbb{R}^{d}, r \in[0, \infty)\right\} be the set of all spheres in Rd\mathbb{R}^{d}. Suppose H={1A:AA}\mathcal{H}=\left\{\mathbf{1}_{A}: A \in \mathcal{A}\right\}. Show that

VC(H)d+2\mathrm{VC}(\mathcal{H}) \leqslant d+2

[\left[\right. Hint: Consider the class of functions F={fc,r:cRd,r[0,)}\mathcal{F}^{\prime}=\left\{f_{c, r}: c \in \mathbb{R}^{d}, r \in[0, \infty)\right\}, where

fc,r(x)=x222cTx+c22r2f_{c, r}(x)=\|x\|_{2}^{2}-2 c^{T} x+\|c\|_{2}^{2}-r^{2}