Free Personalization
Free personalization in machine learning aims to adapt models to individual users without requiring extensive retraining or user-specific data, focusing on efficiency and privacy. Current research emphasizes developing "tuning-free" methods using diffusion models, stochastic optimal control, and generative models to achieve personalization, often leveraging techniques like joint-image distributions or low-rank plus sparse weight decomposition. This area is significant because it enables the creation of customized AI experiences across various applications, from image generation and text-to-image synthesis to personalized dialogue systems and emotion recognition, while mitigating the challenges of data scarcity and privacy concerns associated with traditional personalization approaches.
Papers
FedPC: Federated Learning for Language Generation with Personal and Context Preference Embeddings
Andrew Silva, Pradyumna Tambwekar, Matthew Gombolay
Sample-Efficient Personalization: Modeling User Parameters as Low Rank Plus Sparse Components
Soumyabrata Pal, Prateek Varshney, Prateek Jain, Abhradeep Guha Thakurta, Gagan Madan, Gaurav Aggarwal, Pradeep Shenoy, Gaurav Srivastava