Notations

Notations Explanation
n or N number of observations
D number of variables
d index of the variables in [1,,D]
K number of classes
k index of the classes in [1,,K]
T training set space
X space of variables
X random variable from X
x realization of X
X matrix of variables in Rn×D
\boldsymbol{x}_i column vector of the i^{th} observation
x_{id} scalar of the i^{th} observation and the d^{th} variable of \mathbf{X}
\mathcal{Y} space of response
\mathrm{Y} random variable from \mathcal{Y}
y realization of \mathrm{Y}
\mathbf{y} observed response vector in \mathbb{R}^n
y_i scalar of the i^{th} observation of \mathbf{y}
\eta_i linear predictor of i^{th} observation
f : X \rightarrow Y some fixed but unknown function of \mathbf{X}
s(x) smooth monotonic function of x
||.||_1 \ell_1 norm
||.||_2^2 \ell_2 squared norm
\xi knot
K_d number of knots of variable d
\mathrm{B}(x) B-spline basis functions of x
m polynomial order
\mathbf{B} B-spline basis function matrix in \mathbb{R}^{n \times (k + m +1)}
N(x) natural spline basis functions of x
\mathbf{N} natural splines basis function matrix in \mathbb{R}^{n \times n}
\mathbf{W} penalty matrix of natural splines in \mathbb{R}^{n \times n}
\Omega training set to classify
S number of levels in the hierarchical tree of \Omega
s index of the levels of the hierarchy in [1, \dots, S]
h_s height of the s^{th} level of the hierarchy
\mathcal{G} group partition of all S levels of the hierarchy
G_s number of groups at s^{th} level of the hierarchy
\hat{G}_s^* optimal number of groups
g index of number of groups in [1, \dots, G_s]
\mathcal{G}^s a partition of \Omega at the s^{th} level
\mathcal{G}^s_g g^{th} group of variables at the s^{th} level
\rho_s weights attributed to group partition \mathcal{G}^s
\tilde{\mathbf{X}}^{(s)} matrix of supervariables at s^{th} level of the hierarchy
\tilde{\mathbf{X}}^{(s)}_{\mathcal{G}^s} matrix of supervariables for partition \mathcal{G}^s
\mathbf{X}_{\mathcal{G}} variables matrix of concatenated group partition
\mathbf{X}^s_{\mathcal{G}^s} variables matrix of group partition \mathcal{G}^s
\mathit{G} genomic view
\mathbf{G} matrix of genotype data
\mathbf{Z} matrix of additively coded SNPs
\mathit{M} metagenomic view
\mathbf{M} matrix of metagenomic data
\mathcal{M} group structure of metagenomic data
\boldsymbol{\Delta}_{\mathit{G}\mathit{M}} matrix of interaction terms between genome and metagenome data