Towards Self-Adaptive Cooperative Learning in Collective Systems

Nowadays, collective adaptive systems have become crucial as modern systems increasingly adopt this vision. These systems can be leveraged as a means to facilitate cooperative adaptive learning. However, implementing such systems presents various challenges, including: scalability, failures, non-iid data and complex architectures. This paper presents a modern approach to cooperative and privacy-resilient learning by leveraging macroprogramming. Specifically, we propose a new framework based on the integration of aggregate computing and federated learning, aiming to address these challenges and enhance the effectiveness and security of cooperative learning systems.