IJCAI 19填坑1

Metric-Based Auto-Instructor for Learning Mixed Data Representation

Summary

就是为了解决一般算法不能兼顾特征之间coupling和object之间差异的一种办法
一种思想吧

Research Objective作者的研究目标。

To address these issues, we propose an auto-instructive representation learning scheme to enable margin-enhanced distance metric learning for a discrimination-enhanced representation.

Problem Statement问题陈述，需要解决的问题是什么？

同上

Method(s)作者解决问题的方法/算法是什么？是否基于前人的方法？

Contribution：
A comprehensive representation for mixed data simultaneously learns (1) the couplings between categorical features and continuous features at the feature level, and (2) the discrimination between objects at the object level.

An auto-instructive representation learning scheme with two collaborative instructors learns more discriminative representation between objects by learning the margin- enhanced distance metric.

A metric-based auto-instructor (MAI) model built on two compatible encoding feature spaces is devised to capture the feature-level couplings and enhance the object-level discrimination for the representation of mixed data.

Evaluation作者如何评估自己的方法，实验的setup是什么样的，有没有问题或者可以借鉴的地方。

聚类对比，证明估计的metric还可以。

Conclusion作者给了哪些结论，哪些是strong conclusions, 哪些又是weak的conclusions?

统一了空间，说自己的方法很好

Notes(optional) 不符合此框架，但需要额外记录的笔记。

Reference