Learning Implicit Generative Models with the Method of Learned Moments
This work addresses the challenge of moment estimation in generative modeling for image synthesis, offering a novel training approach with strong performance gains, though it is incremental in the context of existing generative adversarial methods.
The paper tackles the problem of training large-scale implicit generative models by proposing a Method of Learned Moments (MoLM) algorithm, which uses a moment network and asymptotic theory to define and optimize moments, resulting in generators that achieve significantly higher Inception Scores and lower Frechet Inception Distances on CIFAR-10 and nearly perfect Multi-Scale Structural Similarity Scores on CelebA for 128x128 images.
We propose a method of moments (MoM) algorithm for training large-scale implicit generative models. Moment estimation in this setting encounters two problems: it is often difficult to define the millions of moments needed to learn the model parameters, and it is hard to determine which properties are useful when specifying moments. To address the first issue, we introduce a moment network, and define the moments as the network's hidden units and the gradient of the network's output with the respect to its parameters. To tackle the second problem, we use asymptotic theory to highlight desiderata for moments -- namely they should minimize the asymptotic variance of estimated model parameters -- and introduce an objective to learn better moments. The sequence of objectives created by this Method of Learned Moments (MoLM) can train high-quality neural image samplers. On CIFAR-10, we demonstrate that MoLM-trained generators achieve significantly higher Inception Scores and lower Frechet Inception Distances than those trained with gradient penalty-regularized and spectrally-normalized adversarial objectives. These generators also achieve nearly perfect Multi-Scale Structural Similarity Scores on CelebA, and can create high-quality samples of 128x128 images.