從互補標籤學習化約至機率估計

Abstract

互補標籤學習 (Complementary-Label Learning, CLL) 是一個弱監督學習問題，其目標在於僅從互補標籤 (Complementary Labels) 訓練出一個分類器，其中互補標籤指是某個資料「不」屬於的類別。已知方法的主要思想是將此問題化約成一般的分類問題，並設計特殊的轉換以及代理損失函數使互補標籤可以與一般的分類問題連結，但這類的方法卻有一些缺點，例如容易過度擬合。在此論文中，我們設計一個新的框架「化約成互補標籤的分布估計」以避開先前方法可能有的缺點。我們證明了準缺地估計互補標籤的分布再加上一個簡單的解碼即可準確地分類未見過的資料。這個框架更可以解釋一些先前互補標籤學習的重要方法，並使他們在有雜訊的資料集中變得更穩健。此外，這個框架揭示了機率估計的準確度能夠用來驗證模型的準確度。由於此框架以機率估計為基礎，因此不論是深度模型或是傳統方法都能在此框架下進行互補標籤學習。我們同時以實驗驗證此框架在不同情境下皆有一定的準確度以及穩健性。最後，我們也收集、分析並公開了一個由真實人類標記，而非人工生成的互補標籤資料集：CLCIFAR。

Complementary-Label Learning (CLL) is a weakly-supervised learning problem that aims to learn a multi-class classifier from only complementary labels, which indicate a class to which an instance does not belong. Existing approaches mainly adopt the paradigm of reduction to ordinary classification, which applies specific transformations and surrogate losses to connect CLL back to ordinary classification. Those approaches, however, face several limitations, such as the tendency to overfit. In this paper, we sidestep those limitations with a novel perspective--reduction to probability estimates of complementary classes. We prove that accurate probability estimates of complementary labels lead to good classifiers through a simple decoding step. The proof establishes a reduction framework from CLL to probability estimates. The framework offers explanations of several key CLL approaches as its special cases and allows us to design an improved algorithm that is more robust in noisy environments. The framework also suggests a validation procedure based on the quality of probability estimates, offering a way to validate models with only CLs. The flexible framework opens a wide range of unexplored opportunities in using deep and non-deep models for probability estimates to solve CLL. Empirical experiments further verified the framework's efficacy and robustness in various settings. To further analyze the properties of complementary labels in real world, a CIFAR-based complementary dataset, CLCIFAR, was also collected, analyzed, and released publicly.