Bayesian Representation Learning with Oracle Constraints
Theofanis Karaletsos Computational Biology, Sloan Kettering Institute 1275 York Avenue, New York, USA、Serge Belongie Cornell Tech 111 8th Avenue #302, New York, USA、Gunnar Rätsch Computational Biology, Sloan Kettering Institute 1275 York Avenue, New York, USA
Abstract
Representation learning systems typically rely on massive amounts of labeled data in order to be trained to high accuracy. Recently, high-dimensional parametric models like neural networks have succeeded in building rich representations using either compressive, reconstructive or supervised criteria. However, the semantic structure inherent in observations is oftentimes lost in the process. Human perception excels at understanding semantics but cannot always be expressed in terms of labels. Thus, oracles or human-in-the-loop systems, for example crowdsourcing, are often employed to generate similarity constraints using an implicit similarity function encoded in human perception. In this work we propose to combine generative unsupervised feature learning with a probabilistic treatment of oracle information like triplets in order to transfer implicit privileged oracle knowledge into explicit nonlinear Bayesian latent factor models of the observations. We use a fast variational algorithm to learn the joint model and demonstrate applicability to a well-known image dataset. We show how implicit triplet information can provide rich information to learn representations that outperform pre
中文速览
在没有大量标注数据的情况下，如何让机器学习模型真正理解数据的语义结构，是这项工作试图回答的核心问题。研究者提出了一种名为"Oracle优先信念网络"（Oracle-Prioritized Belief Network, OPBN）的联合生成模型，将人类感知相似性判断所产生的三元组约束（triplet constraints）——即"A比C更像B"这类比较信息——与变分自编码器（Variational Autoencoder, VAE）式的无监督特征学习结合起来，通过快速变分推断算法同时从原始观测数据和三元组中学习潜在表示。在人脸图像数据集上的实验表明，这种方法不仅在各类预测任务中优于传统度量学习方法和纯生成模型，还能让潜在空间自动按照语义维度（如光照、身份）分解，使隐变量具有可解释性。这项工作的重要意义在于：它为那些标签稀缺、但可以借助众包等方式廉价获取人类相似性直觉的场景，提供了一条将隐性人类知识转化为可解释显式模型的有效路径。
原文 arXiv:1506.05011;中英对照 + 大白话阅读 https://aha.fim.ai/paper/1506.05011v4