This paper addresses the challenge of balancing diversity and accuracy in recommender systems. The authors introduce a new algorithm, "heat-spreading" (HeatS), designed to enhance diversity in recommendations. HeatS operates on a bipartite user-object network, redistributing "resource" values through an averaging process, which helps in identifying novel and personalized recommendations. The algorithm is then hybridized with a diffusion-based method (ProbS) to combine the strengths of both approaches. By tuning the hybridization parameter, the algorithm can be optimized for either accuracy or diversity, or a balance between the two. The performance of the hybrid algorithm is evaluated using four metrics: recovery of deleted links, precision and recall enhancement, personalization, and surprisal/novelty. The results show that the hybrid algorithm outperforms individual methods in terms of both accuracy and diversity, demonstrating a practical solution to the diversity-accuracy dilemma in recommender systems.This paper addresses the challenge of balancing diversity and accuracy in recommender systems. The authors introduce a new algorithm, "heat-spreading" (HeatS), designed to enhance diversity in recommendations. HeatS operates on a bipartite user-object network, redistributing "resource" values through an averaging process, which helps in identifying novel and personalized recommendations. The algorithm is then hybridized with a diffusion-based method (ProbS) to combine the strengths of both approaches. By tuning the hybridization parameter, the algorithm can be optimized for either accuracy or diversity, or a balance between the two. The performance of the hybrid algorithm is evaluated using four metrics: recovery of deleted links, precision and recall enhancement, personalization, and surprisal/novelty. The results show that the hybrid algorithm outperforms individual methods in terms of both accuracy and diversity, demonstrating a practical solution to the diversity-accuracy dilemma in recommender systems.