15.3 kB
Ctrl+K
Close both named gaps: all four QCQP cases (convex and unknown-f* included) and all three SVM datasets reported as test misclassification error
16904b7 verified - claim-1-algorithm-2-reaches-prescribed-tolerance-with-linear-convergence-in-expectation-on-strongly-convex-smooth-objectives-theorem-4-4
- claim-2-algorithm-3-dows-with-randomized-feasibility-exhibits-the-o-1-sqrt-t-worst-case-horizon-scaling-for-convex-nonsmooth-objectives-theorem-5-3
- claim-3-algorithm-4-t-dows-retains-the-adaptive-rate-without-requiring-bounded-y-theorem-5-5
- claim-4-randomized-feasibility-produces-geometric-infeasibility-reduction-with-the-number-of-updates-lemma-3-1
- claim-5-paper-scale-qcqp-simulations-reproduce-predicted-function-value-and-infeasibility-decay
- claim-6-algorithms-3-and-4-provide-parameter-free-constrained-optimization-without-problem-specific-step-size-tuning
- conclusion
- executive-summary
- 1.74 kB