MintNet: Building Invertible Neural Networks with Masked Convolutions
This work addresses the need for flexible and efficient invertible networks for researchers and practitioners in machine learning, offering incremental improvements in generative modeling and classification.
The authors tackled the problem of constructing invertible neural networks by proposing a new method using masked convolutions and composition rules, resulting in architectures competitive with ResNets on classification tasks and achieving bits per dimension of 0.98, 3.32, and 4.06 on MNIST, CIFAR-10, and ImageNet 32x32 as generative models.
We propose a new way of constructing invertible neural networks by combining simple building blocks with a novel set of composition rules. This leads to a rich set of invertible architectures, including those similar to ResNets. Inversion is achieved with a locally convergent iterative procedure that is parallelizable and very fast in practice. Additionally, the determinant of the Jacobian can be computed analytically and efficiently, enabling their generative use as flow models. To demonstrate their flexibility, we show that our invertible neural networks are competitive with ResNets on MNIST and CIFAR-10 classification. When trained as generative models, our invertible networks achieve competitive likelihoods on MNIST, CIFAR-10 and ImageNet 32x32, with bits per dimension of 0.98, 3.32 and 4.06 respectively.