论文标题
通过表征理想的表征来改善自我监督的学习
Improving Self-Supervised Learning by Characterizing Idealized Representations
论文作者
论文摘要
尽管自我监督学习（SSL）方法取得了经验成功，但尚不清楚其表示的特征导致了高下游精度。在这项工作中，我们表征了SSL表示应该满足的属性。具体而言，我们证明了必要和充分的条件，以至于对于任何任务不变，可以给出的数据增强，以该表示的培训的所需探针（例如，线性或MLP）具有完美的准确性。这些要求导致一个统一的概念框架，用于改进现有的SSL方法并得出新方法。对于对比学习，我们的框架规定了对以前的方法（例如使用不对称投影头）的简单但重大改进。对于非对立学习，我们使用框架来得出一个简单新颖的目标。我们所得的SSL算法在标准基准测试上的表现优于基线，包括Imagenet线性探测的SHAV+多功能。
Despite the empirical successes of self-supervised learning (SSL) methods, it is unclear what characteristics of their representations lead to high downstream accuracies. In this work, we characterize properties that SSL representations should ideally satisfy. Specifically, we prove necessary and sufficient conditions such that for any task invariant to given data augmentations, desired probes (e.g., linear or MLP) trained on that representation attain perfect accuracy. These requirements lead to a unifying conceptual framework for improving existing SSL methods and deriving new ones. For contrastive learning, our framework prescribes simple but significant improvements to previous methods such as using asymmetric projection heads. For non-contrastive learning, we use our framework to derive a simple and novel objective. Our resulting SSL algorithms outperform baselines on standard benchmarks, including SwAV+multicrops on linear probing of ImageNet.
