Pixyz: a Python library for developing deep generative models
This provides a tool for researchers and practitioners to more easily develop and experiment with DGMs, though it is incremental as it builds on existing concepts.
The authors tackled the need for a simple and generic framework to implement deep generative models (DGMs) by developing Pixyz, a Python library that uses step-by-step APIs and memoization to enable concise implementation and faster training, with experimental results showing it outperforms existing probabilistic programming languages in speed.
With the recent rapid progress in the study of deep generative models (DGMs), there is a need for a framework that can implement them in a simple and generic way. In this research, we focus on two features of DGMs: (1) deep neural networks are encapsulated by probability distributions, and (2) models are designed and learned based on an objective function. Taking these features into account, we propose a new Python library to implement DGMs called Pixyz. This library adopts a step-by-step implementation method with three APIs, which allows us to implement various DGMs more concisely and intuitively. In addition, the library introduces memoization to reduce the cost of duplicate computations in DGMs to speed up the computation. We demonstrate experimentally that this library is faster than existing probabilistic programming languages in training DGMs.