Approximation quality for matrix lnsp3937

<p>Approximation quality for matrix lnsp3937. The left y-axis refers to the approximation errors <span class="math-tex">\(e^*\)</span> divided by <span class="math-tex">\(‖A‖_F\)</span> , where <span class="math-tex">\(e^*\)</span> denotes either <span class="math-tex">\(e_{rand}\)</span> (3), <span class="math-tex">\(e_{det}\)</span> (5), or the smallest error possible by the TSVD. The right y-axis refers to the error indicators <span class="math-tex">\(E^*\)</span> divided by <span class="math-tex">\(‖A‖_F\)</span> , where <span class="math-tex">\(E^*\)</span> denotes either <span class="math-tex">\(E_{rand}\)</span> (4) or <span class="math-tex">\(E_{det}\)</span> (10). As mentioned in Section II-A2, the error<br> indicator for RandQB_EI fails in double precision floating point arithmetic for approximation qualities smaller than <span class="math-tex">\(2.1\cdot 10^{-7}\)</span>.</p>