Position-Aware Sequential Attention for Accurate Next Item Recommendations

Timur Nabiev, Evgeny Frolov

Sequential self-attention models usually rely on additive positional embeddings, which inject positional information into item representations at the input. In the absence of posit…