How does target network update frequency affect training stability?

Target network update frequency balances stability and learning speed in DQN.

If updated too frequently, the network becomes unstable because targets shift rapidly. If updated too slowly, learning becomes outdated and sluggish. Soft updates (Polyak averaging) are often used to smooth transitions between networks.