Classification tasks are usually evaluated in terms of accuracy. However, accuracy is discontinuous and cannot be directly optimized using gradient ascent. Popular methods minimize cross-entropy, hinge loss, or other surrogate losses, which can lead to suboptimal results. In this paper, we propose a new optimization framework by introducing stochasticity to a model's output and optimizing expected accuracy, i.e. accuracy of the stochastic model. Extensive experiments on linear models and deep image classification show that the proposed optimization method is a powerful alternative to widely used classification losses.

翻译：分类任务通常以准确率作为评估指标。然而，准确率是一个不连续的函数，无法直接通过梯度上升进行优化。主流方法采用交叉熵损失、合页损失或其他替代损失函数，这可能导致次优结果。本文提出了一种新的优化框架，通过向模型输出引入随机性并优化期望准确率（即随机模型的准确率）。在线性模型和深度图像分类任务上的大量实验表明，所提出的优化方法是广泛使用的分类损失函数的有力替代方案。