Concept超网络 / 超网络方法
HyperNetwork
A neural network architectural approach in which a small auxiliary network (the hypernetwork) dynamically generates the weight parameters of a larger primary network. The technique is commonly applied in conditional generation tasks and parameter-sharing scenarios, enabling flexible weight production without explicitly storing a full separate parameter set for each condition.
Timeline (last 90 days)
Oct 9
A HyperNetwork uses a small network to generate the weights of another larger network, which is related to MaRN's approach but places more emphasis on conditional generation.
Unverified50%