1: \begin{definition} The classifier representing the  random guessing of $K$ different classes is known as the uniformly random classifier (\textsf{RAND}) and it is given by

2: \begin{equation}

3: \label{RAND}

4:  \hat{c}_{R} = \textrm{Unif}\{1,K\}.

5: \end{equation}

6: \noindent where $\textrm{Unif}\{1,K\}$ is the discrete uniformly random function which assigns a categorical class $c_i$ to an example $\mathbf{x}$ with probability $1/K$. All the performance scores contemplated in Table \ref{scores} assign the same value to the goodness of this classifier: 

7:  \begin{eqnarray}

8:  \begin{tabular}{c}

9: $\mathcal{P}^i=\mathcal{R}^i = \frac{1}{K}, \forall i\text{, and }\mathcal{A}cc=\mathcal{A}=\mathcal{G}=\mathcal{H}=max=min=\frac{1}{K}.$\nonumber

10: \end{tabular}

11: \end{eqnarray}

12: In general, for the H\"{o}lder means among the recalls\footnote{This can be trivially proved by following a similar reasoning to Theorem 1 of \cite{Ort16} but using the H\"{o}lder mean equality instead.} (independently of the values of weights, $\boldsymbol{\zeta}$, and the exponent $p$), it holds that

13: $$M_p(\mathbf{R},\boldsymbol{\zeta})= \frac{1}{K}, \text{ where } \mathbf{R}=(\mathcal{R}^1,\ldots,\mathcal{R}^K).$$ 

14: 

15: \end{definition}

16: