EigenDamage: Structured Pruning in the Kronecker-Factored Eigenbasis
This work addresses the need for efficient neural network inference on devices with limited resources, representing an incremental improvement over existing Hessian-based pruning methods.
The paper tackles the problem of reducing neural network size and computational cost for resource-constrained devices by introducing a novel reparameterization in the Kronecker-factored eigenbasis for structured pruning, achieving a 10x reduction in model size and 8x reduction in FLOPs on wide ResNet32 with negligible accuracy loss.
Reducing the test time resource requirements of a neural network while preserving test accuracy is crucial for running inference on resource-constrained devices. To achieve this goal, we introduce a novel network reparameterization based on the Kronecker-factored eigenbasis (KFE), and then apply Hessian-based structured pruning methods in this basis. As opposed to existing Hessian-based pruning algorithms which do pruning in parameter coordinates, our method works in the KFE where different weights are approximately independent, enabling accurate pruning and fast computation. We demonstrate empirically the effectiveness of the proposed method through extensive experiments. In particular, we highlight that the improvements are especially significant for more challenging datasets and networks. With negligible loss of accuracy, an iterative-pruning version gives a 10$\times$ reduction in model size and a 8$\times$ reduction in FLOPs on wide ResNet32.