Enhancing Accuracy and Parameter-Efficiency of Neural Representations for Network Parameterization
This addresses the problem of balancing model performance and computational efficiency for researchers and practitioners in neural network design, though it appears incremental in refining existing parameterization methods.
The paper tackles the trade-off between accuracy and parameter efficiency in neural network weight parameterization, finding that weight reconstruction alone can effectively recover original model accuracy and proposing a novel training scheme that decouples reconstruction from auxiliary objectives to achieve significant improvements over state-of-the-art approaches.
In this work, we investigate the fundamental trade-off regarding accuracy and parameter efficiency in the parameterization of neural network weights using predictor networks. We present a surprising finding that, when recovering the original model accuracy is the sole objective, it can be achieved effectively through the weight reconstruction objective alone. Additionally, we explore the underlying factors for improving weight reconstruction under parameter-efficiency constraints, and propose a novel training scheme that decouples the reconstruction objective from auxiliary objectives such as knowledge distillation that leads to significant improvements compared to state-of-the-art approaches. Finally, these results pave way for more practical scenarios, where one needs to achieve improvements on both model accuracy and predictor network parameter-efficiency simultaneously.