Competitive learning is a learning method of artificial neural networks.

When the network structure is fixed, the learning process is reduced to modifying the connection rights, among which competitive learning refers to the competition among all units in the network unit group for the right to respond to external stimulus patterns.

The connection weight of the unit that wins the competition changes in a direction that is more favorable to the competition of this stimulation pattern. Relatively speaking, the unit that wins the competition suppresses the response of the unit that loses the competition to the stimulation pattern. This adaptive learning enables the network unit to have the characteristic of selecting and accepting external stimulation patterns. The more general form of competitive learning is to allow not only a single winner but multiple winners to appear. Learning occurs on the connection weights of each unit in the winner set.