Fast Flow Reconstruction via Robust Invertible nxn Convolution
This work addresses a bottleneck in flow-based generative models for researchers and practitioners, offering an incremental improvement over existing methods.
The paper tackles the limited flexibility of invertible 1x1 convolutions in flow-based generative models by proposing a novel invertible nxn convolution approach, which improves performance on datasets like CIFAR-10, ImageNet, and Celeb-HQ while using fewer parameters.
Flow-based generative models have recently become one of the most efficient approaches to model data generation. Indeed, they are constructed with a sequence of invertible and tractable transformations. Glow first introduced a simple type of generative flow using an invertible $1 \times 1$ convolution. However, the $1 \times 1$ convolution suffers from limited flexibility compared to the standard convolutions. In this paper, we propose a novel invertible $n \times n$ convolution approach that overcomes the limitations of the invertible $1 \times 1$ convolution. In addition, our proposed network is not only tractable and invertible but also uses fewer parameters than standard convolutions. The experiments on CIFAR-10, ImageNet and Celeb-HQ datasets, have shown that our invertible $n \times n$ convolution helps to improve the performance of generative models significantly.