Decentralized Learning refers to the ability of training a deep/machine learning model without the need of centralizing the raw training data. This is particularly useful in scenarios in which the privacy of such raw data is of paramount importance, and can be helpful in reducing the total network traffic headed to the cloud that is generated by edge devices, such as smartphones and IoT devices. In fact, decentralized learning techniques leave the training data distributed on the devices that have produced the data, while being able to train a model by exchanging ephemeral updates.