multi-objective rec, recommendation systems
Multi-objective recommendation balances competing goals like engagement revenue and user satisfaction simultaneously.
751 technical terms and definitions
Multi-objective recommendation balances competing goals like engagement revenue and user satisfaction simultaneously.
Conversations with multiple participants.
Split design into multiple masks for multi-patterning.
Use multiple litho-etch cycles to achieve finer pitch than single exposure.
Share wafer among multiple customers to reduce prototyping cost.
Share mask costs across designs.
Combine multiple prompts.
Share key/value across heads to reduce memory and speed up inference.
Share K/V across attention heads.
Generate and use multiple queries.
Generate multiple query variations and retrieve with each for broader coverage.
Multi-query strategies generate multiple query variations retrieving for each.
Deploy across geographic regions for resilience.
Hierarchical hash encoding.
Multi-resolution hash encoding stores features at multiple scales in hash tables.
Train on multiple resolutions.
Optimize several responses simultaneously.
Discriminators at different resolutions.
Multi-scale generation produces images at multiple resolutions simultaneously or progressively.
Test at multiple scales.
Process multiple resolutions.
Combine multiple sensors.
Multi-site testing probes multiple die simultaneously increasing throughput by parallelizing measurements across sites.
Multi-skilled operators competently perform various tasks enabling flexibility.
Adapt from multiple source domains.
Multiple checks at different points.
Multi-stage retrieval progressively filters candidates with increasingly expensive methods.
Multi-stakeholder recommendation optimizes for outcomes benefiting multiple parties including users providers and platforms simultaneously.
Balance user provider and platform interests.
Sequential etch steps with different chemistries.
Use multiple steps to gradually bypass restrictions.
Gradually elicit harmful behavior.
Multi-style training uses diverse acoustic conditions during ASR training for robustness.
Adapt to multiple target domains.
Advantages of joint training.
Pre-train on multiple objectives simultaneously.
Multi-task reinforcement learning trains single agent on multiple tasks simultaneously leveraging shared structure.
Train on multiple tasks together.
Learn from multiple teacher models.
Share resources across teams.
Multi-token prediction forecasts several future tokens enabling faster generation.
Back-and-forth dialogue.
Conversation with multiple back-and-forth exchanges.
Multiple supply voltages on chip.
Multi-view learning leverages multiple representations or modalities of data to improve model robustness and performance.
Learn from different views of data.
Dense reconstruction from multiple views.
Multi-threshold designs mix transistors with different threshold voltages optimizing speed and leakage.
Use transistors with different threshold voltages for power/performance trade-offs.
Cell libraries with different thresholds.