Deep Learning for Recommendations

Multi-task Learning for Recommendations

Learning paradigm of simultaneously training multiple objectives (CTR, CVR, session time) to improve generalization and efficiency. MTL shares representations while specializing in specific tasks.

← Back