A partitioning strategy is the algorithm or policy used to divide variables, data, or computation across distributed resources. In distributed machine learning, it can determine how model variables are assigned across parameter servers.
The algorithm by which variables are divided across parameter servers.