Two-Stage Deep Learning - YouTube Video Recommendation Process
Cheepu Balakrishna, D. Sharada Mani, A Basi Reddy, P. Chandrakanth, R. Senthamil Selvan · 2023
YouTube is one of the most widely used and technically advanced commercial systems of recommendations. It provides an overview of the system and highlights the remarkable gains in efficiency made possible by deep learning. This research uses the traditional “two-stage” framework for separating and retrieving information. First, a profound model for generating candidates ought to be described, and then a model for ranking candidates ought to be outlined, and both of these should be done in that order. The order in which these two tasks ought to be completed is important. In addition, it offers instruction that can be put into practice as well as valuable insights that have been gleaned from the process of developing, iterating, and managing a large-scale recommendation system that has a significant impact on the experience of the user.