AI 용어집
인공지능 완전 사전
Contextual Bandit Algorithm
Online learning approach that balances exploration and exploitation by dynamically adapting recommendations according to user context in real-time.
Streaming Implicit Feedback
Continuous capture of user behavioral signals (clicks, viewing time) to instantly adjust recommendations without explicit feedback.
Online Matrix Factorization
Matrix factorization technique updated continuously with each new user interaction to maintain fresh recommendations.
Time-Sensitive Recommendation
Model integrating dynamic temporal factors like time of day, seasonality, or emerging trends to optimize instant relevance.
Adaptive Concept Drift
Automatic detection and correction of changes in user preferences or content popularity to maintain recommendation effectiveness.
Streaming Collaborative Filtering
Streaming version of collaborative filtering processing user interactions as they occur to generate instant recommendations.
Weighted Temporal Similarity
Similarity measure giving more weight to recent interactions to reflect the natural evolution of user preferences.
Adaptive Hybridization
Dynamic combination of multiple recommendation approaches whose weights are continuously adjusted according to real-time performance.
Streaming Sequential Model
Architecture processing user interaction sequences on the fly to predict and recommend the next relevant content.
Continuous Intelligent Exploration
Strategy balancing discovery of new content and exploitation of known preferences to maximize long-term engagement.
Instant Contextual Recommendation
System integrating real-time contextual data (location, weather, device) to immediately personalize suggested recommendations.
Hierarchical Multi-Armed Bandit
Structured approach simultaneously optimizing recommendations at multiple levels (categories then contents) for maximum efficiency.
Distributed Evolutionary Learning
Infrastructure horizontally scaling model updates in parallel to support millions of simultaneous interactions.