Recommender Systems

This chapter first introduces the history of the recommender system and the revolutionary changes in the field of recommender systems. Then, this chapter introduces the basic principles of recommender systems, including introducing the basic assumptions of recommendation algorithms from the perspective of machine learning, introducing how to define the recommendation problem in the form of a machine learning problem, and emphatically introducing the deep learning-based paradigm to solve the recommendation problem—“representation learning + interaction function learning”. This chapter also gives an overview of the technical architecture of recommender systems, including the differences between small- and medium-scale recommender systems and large-scale recommender systems. Finally, this chapter introduces the main application areas of recommender systems, such as e-commerce, content platforms, etc., and the actual business value brought by recommender systems to these application areas and compares the three main applications in the Internet field—search, advertising, and recommendation, by the differences and connections among them. Starting from industry problems, this chapter summarizes the differences in the application of recommender systems in different industries and outlines the solutions to different types of problems.

This chapter introduces four types of classic recommendation algorithms, including content-based recommendation algorithms, classic collaborative filtering algorithms, matrix factorization methods, and factorization machines. Before the emergence of deep learning, these methods were the most mainstream techniques for recommender systems, widely recognized by both academia and industry. Although after the emergence of deep learning, these technologies are no longer the first choice of the industry, but the basic ideas and practical experience extracted from these technologies still affect the follow-up research. Therefore, in many deep learning-based recommendation algorithms, we can often see the reflections of the above approaches.

This chapter introduces the basics of deep learning, including feedforward computation and backpropagation algorithms for deep neural networks, as well as various classic neural network models. As readers learn, they can combine the content of other chapters in this book to understand and design different types of neural network models for recommendation scenarios, taking into account the data characteristics and task properties, in order to improve recommendation performance.

This chapter introduces the relationship between collaborative filtering and deep learning and then presented various deep learning-based collaborative filtering algorithms. Leveraging cutting-edge methods from deep learning, these algorithms can significantly improve the accuracy, scalability, diversity, and interpretability of recommendation systems, offering richer technological choices for recommendation system design. However, most of these algorithms are optimized for specific problems, and there are often limitations in practical applications. Therefore, at the system design level, algorithm integration or fusion needs to be considered.

This chapter introduces the hotspots of recommender system research, the key challenges of recommender system application, and how to achieve responsible recommendation technically. These contents may become the key of recommender system research and application in the future, so they need the continuous attention of researchers and developers.

This chapter focuses on some problems and considerations in industry applications of recommender systems, and discusses the details of these actual applications based on the code in the Microsoft Recommenders repository and a cloud-based reference architecture. Readers are encouraged to follow the steps and methods in the text in a hands-on manner and experiment using the algorithms introduced in previous chapters.

This chapter provides a summary of the book and offers insights into future trends in the research and application of recommender systems.