Editing Recommender system (section)

=== Generative recommenders ===
Generative recommenders (GR) represent an approach that transforms recommendation tasks into [[Seq2seq|sequential transduction]] problems, where user actions are treated like tokens in a generative modeling framework. In one method, known as HSTU (Hierarchical Sequential Transduction Units),<ref>{{cite arXiv |last1=Zhai |first1=Jiaqi |title=Actions Speak Louder than Words: Trillion-Parameter Sequential Transducers for Generative Recommendations |date=2024-05-06 |eprint=2402.17152 |last2=Liao |first2=Lucy |last3=Liu |first3=Xing |last4=Wang |first4=Yueming |last5=Li |first5=Rui |last6=Cao |first6=Xuan |last7=Gao |first7=Leon |last8=Gong |first8=Zhaojie |last9=Gu |first9=Fangda|class=cs.LG }}</ref> high-[[cardinality]], non-stationary, and streaming datasets are efficiently processed as sequences, enabling the model to learn from trillions of parameters and to handle user action histories orders of magnitude longer than before. By turning all of the system’s varied data into a single stream of tokens and using a custom [[Attention (machine learning)|self-attention]] approach instead of [[Neural network (machine learning)|traditional neural network layers]], generative recommenders make the model much simpler and less memory-hungry. As a result, it can improve recommendation quality in test simulations and in real-world tests, while being faster than previous [[Transformer (deep learning architecture)|Transformer]]-based systems when handling long lists of user actions. Ultimately, this approach allows the model’s performance to grow steadily as more computing power is used, laying a foundation for efficient and scalable “[[Foundation model|foundation models]]” for recommendations.