Search papers, labs, and topics across Lattice.
This study compares pretrained item embeddings derived from graph neural networks (GNNs) with end-to-end trainable embeddings in two large-scale recommendation systems at Yandex, specifically Yandex Market and Yandex Music. The research reveals that while a separate pretraining phase is beneficial in low-data scenarios, it does not significantly enhance performance for large-scale models trained on extensive datasets. These findings provide critical insights into the trade-offs between embedding strategies in industrial applications, informing future model design choices.
Pretraining item embeddings can be a double-edged sword鈥攂eneficial in low-resource settings but redundant in large-scale scenarios.
Transformer-based sequential recommendation models, which process sequences of user-item interactions, rely heavily on the item embedding strategy. Existing approaches either use pretrained item embeddings or learn them end-to-end with the transformer. To the best of our knowledge, no prior work has compared these options from both cost and quality perspectives in a large-scale industrial setting. This paper is a case study that compares pretrained industrial graph neural network item embeddings with end-to-end trainable item embeddings across two mature production recommendation systems at Yandex: Yandex Market and Yandex Music. We additionally evaluate both approaches on a low-resource dataset sampled from Yandex Lavka production logs, for which both the data and code are publicly available for demonstration purposes. Our results show that a separate pretraining stage helps when training data is limited, but provides no worthwhile benefit for large-scale models trained on extensive datasets.
