Mixture of Experts - Graph View

A neural network architecture that uses a gating network to route inputs to specialized sub-networks called experts, enabling efficient scaling by activating only a subset of parameters for each input.