SQS Framework for Simultaneous Pruning and Quantization

SQS uses Bayesian variational learning with a spike-and-slab prior and Gaussian Mixture Models to achieve higher neural network compression rates than individual pruning or quantization methods.