AI-ordlista
Den kompletta ordlistan över AI
Model Personalization
Process of adapting a generic global model to meet the specific preferences and individual characteristics of each user or device.
Local Fine-tuning
Technique involving fine-tuning the parameters of the global model on each client's local data after initial federated training to maximize individual relevance.
Federated Meta-learning
Approach combining meta-learning and federated learning where the model learns to quickly adapt to new tasks or specific preferences of each user.
Weighted Aggregation
Method of fusing local model updates where each contribution is weighted according to quality metrics, data quantity, or relevance to the end user.
Client Adaptation
Strategy where each client performs customized adjustments to the received model according to its specific characteristics before or after global aggregation.
Global Base Model
Fundamental model architecture shared among all clients, serving as a common starting point for local personalized adaptations.
Local Overfitting
Phenomenon where a personalized model over-adapts to a user's specific data, losing its ability to generalize to new inputs.
Data Heterogeneity
Statistical and distributional variation of data between different clients in a federated system, requiring adaptive personalization strategies.
Federated Transfer Learning
Combination of transfer learning and federated learning to effectively adapt global knowledge to the specific contexts of each user.
Federated Multi-task
Extension of federated learning where a single model simultaneously learns multiple customized tasks for different users or contexts.
Distributed Continual Learning
Paradigm where models continuously adapt to new local data without forgetting previous knowledge, while maintaining global consistency.
Federated Clustering
Technique for grouping clients into clusters based on their data similarities or preferences to optimize model personalization by group.
Hybrid Architecture
Combination of multiple personalization approaches (fine-tuning, meta-learning, etc.) in a federated system to optimize adaptation to individual needs.
Local-Global Optimization
Iterative process of balancing between improving customized local performance and maintaining the consistency and global performance of the federated model.