提供相关广告背后的机器学习

Felix Fang | Software Engineer, Advertiser Solutions Group

Felix Fang | 软件工程师，广告商解决方案组

Chi Xu | Software Engineer, Advertiser Solutions Group

徐驰｜软件工程师，广告商解决方案组

Pinterest is where people go to plan and shop, making ideas and ads from brands helpful in taking Pinners from inspiration to action. It’s our goal to ensure ads continue to be additive and not intrusive on Pinterest. Because of the unique and powerful first party signals on the platform, advertisers can reach Pinners based on their interests, intent and engagement on the platform.

Pinterest是人们计划和购物的地方，这使得来自品牌的想法和广告有助于将品客从灵感带到行动。我们的目标是确保广告在Pinterest上继续是附加的，而不是干扰性的。由于平台上独特而强大的第一方信号，广告商可以根据品客的兴趣、意图和在平台上的参与度来接触他们。

To help in delivering the right ads to the right Pinners in an audience of hundreds of millions of people, we offer advertisers features to achieve relevance including Actalike (AAL) audiences, also known in the industry as Lookalike audiences. AAL audiences help advertisers reach potentially new users via audience expansion.

为了帮助在数以亿计的受众中向正确的Pinners提供正确的广告，我们为广告商提供了实现相关性的功能，包括AALike（AAL）受众，在业界也被称为Lookalike受众。AAL受众可以帮助广告商通过受众扩展来接触潜在的新用户。

In this blog, we’ll focus on the machine learning model component of relevant ads delivery and explain how we achieve high quality audience expansion through universal user embedding representations together with per-advertiser classifier models. We demonstrate the power of the proposed combined approach by showing better performance over both regression-based and similarity-based approaches.

在这篇博客中，我们将专注于相关广告投放的机器学习模型部分，并解释我们如何通过通用的用户嵌入表示与每个广告商的分类器模型一起实现高质量的受众扩展。我们通过展示比基于回归和基于相似性的方法更好的性能来证明所提出的组合方法的力量。

Our Proposed Approach: User Embeddings + MLP Classifiers

我们建议的方法。用户嵌入+MLP分类器

AAL methods mainly fall into two categories: regression-based and similarity-based approaches [1, 2, 3, 4, 5, 6]. Regression-based approaches treat the task as a binary classification problem, train a model for each seed list offline, and use such models to score candidate users directly, as shown in Figure 2a. Similarity-based approaches learn...