Extensions of regret-minimization algorithm for optimal design
This work addresses sample selection for experimental design, particularly in image classification, but is incremental as it builds on an existing regret minimization framework.
The authors tackled the problem of optimal experimental design by extending a regret minimization framework with entropy regularization, achieving a provable sample complexity bound for (1+ε)-near optimal solutions. They applied this to select representative samples from image datasets without labels, showing consistent performance improvements over baselines on MNIST, CIFAR-10, and ImageNet subsets.
We explore extensions and applications of the regret minimization framework introduced by~\cite{design} for solving optimal experimental design problems. Specifically, we incorporate the entropy regularizer into this framework, leading to a novel sample selection objective and a provable sample complexity bound that guarantees a $(1+ε)$-near optimal solution. We further extend the method to handle regularized optimal design settings. As an application, we use our algorithm to select a small set of representative samples from image classification datasets without relying on label information. To evaluate the quality of the selected samples, we train a logistic regression model and compare performance against several baseline sampling strategies. Experimental results on MNIST, CIFAR-10, and a 50-class subset of ImageNet show that our approach consistently outperforms competing methods in most cases.