Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
Queen Mary University of London
2
0
2
Sparse embeddings can significantly outperform dense vectors in cold-start recommendations while slashing storage costs.
Content-based cold-start recommendation gets a boost from sparse contrastive learning, showing that selectively ignoring uninformative negative samples sharpens item relevance estimation and beats standard softmax.