4 problems

Let $R$ be the original algorithm set, and introduce auxiliary algorithms into the benchmarking set. Auxiliary-algorithm conjecture. The more auxiliary algorithms are introduced, t…

Let $n$ players be ranked under the Bradley–Terry–Luce model, with skill parameters $\theta^*$, sampling probability $p$, comparison count $L$, permutation space $\mathfrak{S}_n$,…

Let $H(\theta^*)$ be the Hessian matrix of the objective at the true skill parameter $\theta^*$, and let $e_j$ denote the $j$th canonical vector. For players with true ranks…

The BTL model is parameterized by a score vector $\theta^*$, identifiable only up to an additive constant; impose $\mathds{1}_{n}^T\widehat{\theta}=0$ and…