Industrial Recommender System: Principles, Technologies and Enterprise Applications
- Добавил: literator
- Дата: 2-06-2024, 13:56
- Комментариев: 0
![](/uploads/posts/2024-06/thumbs/1717325787_industrial-recommender-system.jpg)
Автор: Lantao Hu, Yueting Li, Guangfan Cui, Kexin Yi
Издательство: Springer/Publishing House of Electronics Industry
Год: 2024
Страниц: 256
Язык: английский
Формат: pdf (true), epub
Размер: 54.0 MB
Recommender systems, as a highly popular Artificial Intelligence (AI) technology in recent years, have been widely applied across various industries. They have transformed the way we interact with technology, influencing our choices and shaping our experiences. This book provides a comprehensive introduction to industrial recommender systems, starting with the overview of the technical framework, gradually delving into each core module such as content understanding, user profiling, recall, ranking, re-ranking and so on, and introducing the key technologies and practices in enterprises. The book also addresses common challenges in recommendation cold start, recommendation bias and debiasing. Additionally, it introduces advanced technologies in the field, such as Reinforcement Learning, causal inference.
Professionals working in the fields of recommender systems, computational advertising, and search will find this book valuable. It is also suitable for undergraduate, graduate, and doctoral students majoring in Artificial Intelligence, Computer Science, software engineering, and related disciplines. Furthermore, it caters to readers with an interest in recommender systems, providing them with an understanding of the foundational framework, insights into core technologies, and advancements in industrial recommender systems.
Industrial online recommender systems not only involve recall, pre-ranking, ranking, and re-ranking but also involve content understanding, user profiles, AB testing platforms, Session context management, producer ecosystem, traffic operation platforms, etc. These subsystems are less mentioned in similar books, where the evolution of model technology often takes up the most space. Taking content understanding as an example, labeling content with tags or obtaining a vector representation through unsupervised learning are two different but useful approaches. Tags can play a significant role in user cold start, while vectorization can be applied to both the cold start of users and content.
The AB testing platform is crucial for recommendation algorithm engineers to iterate strategies. How to design a good experiment and interpret whether the experiment results are significant is a difficult task for recommendation algorithm engineers. To analysis of the experiment requires a foundation in statistical theory and understanding confidence levels, p-values, and other statistical measures.
This book provides a detailed introduction to the important components of the recommendation system from the perspective of an algorithm engineer with several years of experience in industrial recommendation system development. It delves into aspects such as the tag system, user profiles, multimodal content understanding, practical skills for optimizing effects, and other areas that are often overlooked in similar recommendation system books. It also provides clear explanations in conjunction with specific practical scenarios. How to evaluate the quality of a recommendation system is a challenging issue with many perspectives. Whether just focusing on online consumption metrics or constructing a complex multi-level matric matrix, the trade-offs are difficult. This requires combining with the business scenario you are working on and tightly integration with the product or operations team.
Tag extraction has played an important role in the history of recommendation algorithms, widely adopted for its characteristics such as transparency, easy control, and good integration with operational domain knowledge. With the application of Deep Learning technology, from the perspective of metric optimization alone, tags seem to be an outdated technology. However, recommendation cold start is still a challenge that every industrial recommendation system cannot overcome. Cold start algorithms based on user tags combined with E&E (Exploration and Exploitation) strategies or with reinforcement learning can achieve excellent results on this classic difficult problem.
The style of this book is very pragmatic, making it highly suitable for engineers who want to learn about recommender systems, and also suitable for scholars and students engaged in recommendation system research. It provides a comprehensive introduction of industrial recommender systems. We look forward to more excellent technical professionals opening the door to intelligent recommender systems.
Contents:
Скачать Industrial Recommender System: Principles, Technologies and Enterprise Applications
![](/templates/LitmyL/images/LitMyru.png)
[related-news] [/related-news]
Внимание
Уважаемый посетитель, Вы зашли на сайт как незарегистрированный пользователь.
Мы рекомендуем Вам зарегистрироваться либо войти на сайт под своим именем.
Уважаемый посетитель, Вы зашли на сайт как незарегистрированный пользователь.
Мы рекомендуем Вам зарегистрироваться либо войти на сайт под своим именем.