Machine learning system and machine learning method

Machine learning system comprising a host device and a plurality of client devices. The host device is configured to store and update a host model. The client devices are configured to be communicatively coupled to the host device and receive the host model, and comprise a first and second client device, storing a first and a second parameter set respectively and configured to train the received host model according to the first and the second parameter sets, respectively, to respectively generate a first and a second training result. If the host device receives the first training result corresponding to an mth training round but does not receive the second training result corresponding to a nth training round, when a difference between m and n is not larger than a threshold value, and m is larger than n, the host device updates the host model according to the first training result without using the second training result. Also disclosed are another machine learning system and a machine learning method.