Training High-Quality Catalog Item Embeddings with Triplet Loss and Siamese Neural Networks
Source: Deephub Imba This article is about 4500 words long and is recommended for a 5-minute read. This article describes a method for training high-quality, transferable embeddings using self-supervised learning techniques on user search data from within the website. The number of categories in large websites is vast, and manual tagging is generally unfeasible. Therefore, … Read more
