Treffer: Denoising Diffusion Model-driven Adaptive Estimation of Distribution Algorithm Integrating Multi-modal Data

Personalized search and recommendation algorithms for multi-modal data have attracted widespread attention. In view of the insufficient ability of multi-source information fusion and global search in complex optimization problems, this paper proposed a denoising diffusion model-driven adaptive estimation of distribution algorithm integrating multi-modal data. Firstly, multi-modal information is extensively collected from various sources. By introducing diffusion model, a user interest preference model based on the denoising diffusion model is established to extract the user preference features, which fits users’ cognitive experiences and behavioral patterns. Meanwhile, in the framework of an estimation of distribution algorithm, some estimation of distribution strategies are designed to enhance the global exploration and local development ability of the optimization algorithm. Finally, according to new data information, a dynamic model update mechanism is established to promptly update the user interest preference model and related models, which tracks the changes of users’ interest preferences for actual scenarios. It will help users quickly filter out items that match their interest preferences from a vast amount of information. The feasibility, effectiveness and superiority of the proposed algorithm were verified through many experiments on general public datasets. It improved the search efficiency and recommendation effect of the personalized recommendation algorithm.