Meipai AI Tech Talk: Deep Ranking Models, Video Clustering, and Optimization

This article introduces three technical presentations from a Meipai AI tech meetup.

Topic 1: Meipai Personalized Deep Ranking Model - Factorization based Deep Models for Personalized Intelligence

This presentation explores how to use Deep Neural Networks (DNN) to deeply understand user interest preferences and select content from the massive content ecosystem for users. The speaker discusses optimizing deep ranking models to build a low-latency personalized service that is both "deep" enough to fully understand user behavior patterns and capture personalization, while also being "fast" to ensure efficient model inference and avoid unreasonable latency responses.

Topic 2: Meipai Video Clustering Practice

In personalized recommendation systems, understanding user interests through item attributes is common. This presentation introduces a user behavior-based video clustering approach that uses user click and playback behaviors to cluster video content. This method breaks through the limitation of understanding video content from a visual perspective and can discover non-human-curated classification knowledge, thereby improving personalized recommendation effectiveness.

Topic 3: Application of Data Structures in Optimization Problems

Optimization problems are ubiquitous in life, such as minimizing travel time or maximizing investment returns. As data scale expands and consideration factors increase (uncertainty, business factors), solution time also increases significantly. This report introduces common optimization problems and solution methods through two business cases, then proposes new objective functions from an economic perspective, and discusses in detail how data structures accelerate the solving process.