FedPara: Low-rank Hadamard Product Parameterization for Efficient Federated Learning
Efficient federated communication via low-rank Hadamard product decomposition of weight matrices
#federated-learning
#weight-matrix
#low-rank
Recurrent Parameter Generators
Generating parameters from fixed weight matrix promotes efficiency
#model-compression
#weight-matrix
Orthogonalizing Convolutional Layers with the Cayley Transform
Explicit orthogonality of convolution is obtained by Cayley transform
#orthogonality
#weight-matrix