An ideal learned representation should display transferability and robustness. Supervised contrastive learning (SupCon) is a promising method for training accurate models, but produces representations that do not capture these properties due to class collapse -- when all points in a class map to the same representation. Recent work suggests that "spreading out" these representations improves them, but the precise mechanism is poorly understood. We argue that creating spread alone is insufficient for better representations, since spread is invariant to permutations within classes. Instead, both the correct degree of spread and a mechanism for breaking this invariance are necessary. We first prove that adding a weighted class-conditional InfoNCE loss to SupCon controls the degree of spread. Next, we study three mechanisms to break permutation invariance: using a constrained encoder, adding a class-conditional autoencoder, and using data augmentation. We show that the latter two encourage clustering of latent subclasses under more realistic conditions than the former. Using these insights, we show that adding a properly-weighted class-conditional InfoNCE loss and a class-conditional autoencoder to SupCon achieves 11.1 points of lift on coarse-to-fine transfer across 5 standard datasets and 4.7 points on worst-group robustness on 3 datasets, setting state-of-the-art on CelebA by 11.5 points.
翻译:理想的学识代表应该显示可转移性和稳健性。 监督对比学习( SupCon) 是培训准确模型的一个很有希望的方法, 但是由于阶级分布图中的所有点都显示不出这些属性 -- -- 当阶级分布图中的所有点都显示在同一代表面时, 监督对比学习( SupCon) 是一个很有希望的方法, 但是产生由于阶级崩溃而不能捕捉这些属性的演示。 最近的工作表明, “ 扩展” 这些表达面可以改善它们, 但准确的机制却不怎么理解。 我们争论说, 单独创造扩散不足以更好地表达, 因为扩散在阶级内部差异中是不可变异的。 相反, 需要正确的传播程度和打破这种变异的机制。 我们首先证明, 为SupConConE损失添加一个加权的等级条件级信息损失, 并且为SupCelcoder 控制着传播的程度。 接下来, 我们研究三个机制来打破变异变机制: 使用一个受约束的编码器, 添加一个等级自动调节的自动编码, 并使用数据扩增。 我们表明, 后两个机制鼓励在比前者更现实的条件下将潜的亚类分级分级分级分级的分数组在比前者更现实的条件下进行。 我们显示, 在最高级的Celec- 5Celeval- AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS 5 AS AS AS 上, 在 AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS 5 5 5 5 AS AS AS AS AS AS AS AS AS AS 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 调 点上, 调 调