Problems using deep generative models for probabilistic audio source separation
arXiv:2011.01761
Abstract
Recent advancements in deep generative modeling make it possible to learn prior distributions from complex data that subsequently can be used for Bayesian inference. However, we find that distributions learned by deep generative models for audio signals do not exhibit the right properties that are necessary for tasks like audio source separation using a probabilistic approach. We observe that the learned prior distributions are either discriminative and extremely peaked or smooth and non-discriminative. We quantify this behavior for two types of deep generative models on two audio datasets.
References in corpus (8)
- Adam: A Method for Stochastic Optimization
- PyTorch: An Imperative Style, High-Performance Deep Learning Library
- WaveNet: A Generative Model for Raw Audio
- Density estimation using Real NVP
- Amortised MAP Inference for Image Super-resolution
- FloWaveNet : A Generative Flow for Raw Audio
- Source Separation with Deep Generative Priors
- Audio Source Separation via Multi-Scale Learning with Dilated Dense U-Nets