提出量化选择迁移学习数据集的方法，提升小样本图像分类效果
TLDChoiceNet: Quantitatively Choosing a Transfer Learning Dataset
- 构建TLDChoiceNet模型预测微调后准确率，指导数据集选择
- 引入分布距离与类别相关性指标，实现0.97的R2预测精度
- 证明低层统计特征可解释迁移效果，适合小样本研究者使用
在训练数据有限的场景下，迁移学习尤为重要。图像分类中常以ImageNet、CIFAR-100或COCO等大规模数据集作为迁移源。虽然直观上认为迁移数据集应包含更多类别和每类更多样本，但目前尚无定量方法选择最优迁移数据集。本文设计TLDChoiceNet，通过预测微调后的测试集准确率，为给定微调数据集选择最佳迁移数据集。简单版本在测试集上达到0.154 MSE，而利用预训练ResNet50 v2嵌入并加入每类信息的版本将误差降低至0.031（5倍提升）。进一步提出两个无监督评估指标：分布距离（DD）与平均类别相关性（ACC），前者与微调准确率呈线性关系（R²=0.89），后者将R²提升至0.97。结果表明，数据集的低层统计特征可有效解释迁移学习效果，且预训练ImageNet能将不同类别在隐空间中分得更远。
原文摘要 · Abstract (English)
Transfer learning is particularly useful in settings with limited training data, and within image classification it is common to transfer learn upon massive datasets like ImageNet , CIFAR-100, or COCO . Qualitatively, it seems a transfer learning dataset should have both more classes and more examples per class than the fine tuning dataset; however, a quantitative method to choose the best transfer learning dataset does not currently exist. In this paper, we design TLDChoiceNet, a model to choose the best transfer learning dataset given a fine tuning dataset by predicting the test-set accuracy after fine-tuning. A simple version 1 achieves 0.154 MSE on the test dataset, while a version 2 leveraging an ImageNet pre-trained ResNet50 v2 embedding with per-class information attains a 5X lower MSE of 0.031. We further design two metrics that enable an unsupervised method of choosing an optimal transfer learning dataset: distribution distance (DD), which linearly regresses against fine-tune accuracy with an R2 of 0.89, and average class correlation (ACC), which improves the R2 to 0.97. Our results underscore that a dataset's low-level statistics can explain the transfer learning effect, and that using a pre-trained ImageNet can embed different classes further apart in latent feature space.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。