AI 中文总结
该研究针对卷积神经网络冗余特征图问题，提出用多臂赌博机的损失感知特征图剪枝框架，将候选特征图视为臂进行评估，经试验预算后按分数排序移除，评估多种算法，结果显示 UCB1 和汤普森采样能在减计算时保持准确率，优于其他剪枝方法。
AI 中文摘要
卷积神经网络通常包含冗余特征图，增加存储和推理成本。本文提出了一种使用多臂赌博机的损失感知特征图剪枝框架。特征图剪枝是结构化的，因为它移除完整的卷积输出通道及其生成滤波器。每个候选特征图被视为一个臂。每次试验时，一个图被临时屏蔽并在采样小批次上评估；然后恢复该图，并将观察到的损失变化转换为安全移除奖励。在固定试验预算后，候选图按学习分数排序，前 k 个图及其滤波器、偏差和相应的下一层输入通道内核被永久移除。研究评估了 UCB1 和汤普森采样，在 LeNet/MNIST 上与直接/预言风格评估进行比较，并将评估扩展到 MNIST、CIFAR-10、CIFAR-100、SVHN、CUB-200-2011 和牛津花卉 102。结果表明，UCB1 和汤普森采样在移除特征图和减少卷积计算的同时，保持了接近未剪枝模型的准确率。弗里德曼和内门尼检验表明，UCB1 获得最高平均排名，其次是汤普森采样；两者均显著优于贪婪和基于幅度的剪枝，同时在统计上与原始未剪枝模型相当。
英文摘要
Convolutional neural networks often contain redundant feature maps that increase storage and inference cost. This paper presents a loss-aware feature-map pruning framework using multi-armed bandits. Feature-map pruning is structured because it removes complete convolutional output channels and their producing filters rather than isolated scalar weights. Each candidate feature map is treated as an arm. At each play time, one map is temporarily masked and evaluated on a sampled mini-batch; the map is then restored and the observed loss change is converted into a safe-removal reward. After a fixed play budget, candidate maps are ranked by learned scores and the top-k maps are permanently removed with their filters, biases and corresponding next-layer input-channel kernels. The study evaluates UCB1 and Thompson Sampling, compares them with direct/oracle-style evaluation on LeNet/MNIST, and extends the evaluation to MNIST, CIFAR-10, CIFAR-100, SVHN, CUB-200-2011 and Oxford Flowers 102. Results show that UCB1 and Thompson Sampling preserve accuracy close to unpruned models while removing feature maps and reducing convolutional computation. Friedman and Nemenyi tests show that UCB1 obtains the highest mean rank, followed by Thompson Sampling; both significantly outperform greedy and magnitude-based pruning while remaining statistically comparable to the original unpruned model.
Comments18 pages, 4 figures
