Learning Deep Structured Models
Liang-Chieh Chen ∗ Department of Computer Science, UCLA,、Alexander G. Schwing Department of Computer Science, University of Toronto,、Alan L. Yuille Department of Statistics, UCLA,、Raquel Urtasun Department of Computer Science, University of Toronto, The first two authors contributed equally to this work.
Abstract
Many problems in real-world applications involve predicting several random variables which are statistically related. Markov random fields (MRFs) are a great mathematical tool to encode such relationships. The goal of this paper is to combine MRFs with deep learning algorithms to estimate complex representations while taking into account the dependencies between the output random variables. Towards this goal, we propose a training algorithm that is able to learn structured models jointly with deep features that form the MRF potentials. Our approach is efficient as it blends learning and inference and makes use of GPU acceleration. We demonstrate the effectiveness of our algorithm in the tasks of predicting words from noisy images, as well as multi-class classification of Flickr photographs. We show that joint learning of the deep features and the MRF parameters results in significant performance gains.
中文速览
真实应用中往往需要同时预测多个相互关联的变量，比如一张图里同时识别多个单词，而现有深度学习方法通常先单独训练特征提取器、再接结构化预测模型，这种两阶段做法因为忽略了变量间依赖关系而效果有限。本文提出将马尔可夫随机场（Markov Random Field, MRF）与深度神经网络融合在一起端到端联合训练——把深度网络输出直接作为MRF的势函数，同时利用对偶变换把"先推断、再学习"的双层循环转化为单层迭代，让参数更新与消息传递交替进行，并充分利用GPU加速。在从噪声图像中识别单词和Flickr图像多标签分类两个任务上，联合训练相比分阶段训练带来了显著的性能提升。这项工作的价值在于，它为在任意图结构上将深度特征学习与结构化预测无缝结合提供了一套高效可行的框架，让模型能够在学习特征的同时感知输出变量之间的统计依赖，从而做出更准确的联合预测。
原文 arXiv:1407.2538;中英对照 + 大白话阅读 https://aha.fim.ai/paper/1407.2538v3