English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

A large catalogue size is one of the central challenges in trainingrecommendation models: a large number of items makes them memory andcomputationally inefficient to compute scores for all items during training,forcing these models to deploy negative sampling. However, negative samplingincreases the proportion of positive interactions in the training data, andtherefore models trained with negative sampling tend to overestimate theprobabilities of positive interactions a phenomenon we call overconfidence.While the absolute values of the predicted scores or probabilities are notimportant for the ranking of retrieved recommendations, overconfident modelsmay fail to estimate nuanced differences in the top-ranked items, resulting indegraded performance. In this paper, we show that overconfidence explains whythe popular SASRec model underperforms when compared to BERT4Rec. This iscontrary to the BERT4Rec authors explanation that the difference in performanceis due to the bi-directional attention mechanism. To mitigate overconfidence,we propose a novel Generalised Binary Cross-Entropy Loss function (gBCE) andtheoretically prove that it can mitigate overconfidence. We further propose thegSASRec model, an improvement over SASRec that deploys an increased number ofnegatives and the gBCE loss. We show through detailed experiments on threedatasets that gSASRec does not exhibit the overconfidence problem. As a result,gSASRec can outperform BERT4Rec (e.g. +9.47% NDCG on the MovieLens-1M dataset),while requiring less training time (e.g. -73% training time on MovieLens-1M).Moreover, in contrast to BERT4Rec, gSASRec is suitable for large datasets thatcontain more than 1 million items.

gSASRec: Reducing Overconfidence in Sequential Recommendation Trained with Negative Sampling