-
Gotta Go Fast When Generating Data with Score-Based Models
Authors:
Alexia Jolicoeur-Martineau,
Ke Li,
Rémi Piché-Taillefer,
Tal Kachman,
Ioannis Mitliagkas
Abstract:
Score-based (denoising diffusion) generative models have recently gained a lot of success in generating realistic and diverse data. These approaches define a forward diffusion process for transforming data to noise and generate data by reversing it (thereby going from noise to data). Unfortunately, current score-based models generate data very slowly due to the sheer number of score network evalua…
▽ More
Score-based (denoising diffusion) generative models have recently gained a lot of success in generating realistic and diverse data. These approaches define a forward diffusion process for transforming data to noise and generate data by reversing it (thereby going from noise to data). Unfortunately, current score-based models generate data very slowly due to the sheer number of score network evaluations required by numerical SDE solvers.
In this work, we aim to accelerate this process by devising a more efficient SDE solver. Existing approaches rely on the Euler-Maruyama (EM) solver, which uses a fixed step size. We found that naively replacing it with other SDE solvers fares poorly - they either result in low-quality samples or become slower than EM. To get around this issue, we carefully devise an SDE solver with adaptive step sizes tailored to score-based generative models piece by piece. Our solver requires only two score function evaluations, rarely rejects samples, and leads to high-quality samples. Our approach generates data 2 to 10 times faster than EM while achieving better or equal sample quality. For high-resolution images, our method leads to significantly higher quality samples than all other methods tested. Our SDE solver has the benefit of requiring no step size tuning.
△ Less
Submitted 28 May, 2021;
originally announced May 2021.
-
Adversarial score matching and improved sampling for image generation
Authors:
Alexia Jolicoeur-Martineau,
Rémi Piché-Taillefer,
Rémi Tachet des Combes,
Ioannis Mitliagkas
Abstract:
Denoising Score Matching with Annealed Langevin Sampling (DSM-ALS) has recently found success in generative modeling. The approach works by first training a neural network to estimate the score of a distribution, and then using Langevin dynamics to sample from the data distribution assumed by the score network. Despite the convincing visual quality of samples, this method appears to perform worse…
▽ More
Denoising Score Matching with Annealed Langevin Sampling (DSM-ALS) has recently found success in generative modeling. The approach works by first training a neural network to estimate the score of a distribution, and then using Langevin dynamics to sample from the data distribution assumed by the score network. Despite the convincing visual quality of samples, this method appears to perform worse than Generative Adversarial Networks (GANs) under the Fréchet Inception Distance, a standard metric for generative models.
We show that this apparent gap vanishes when denoising the final Langevin samples using the score network. In addition, we propose two improvements to DSM-ALS: 1) Consistent Annealed Sampling as a more stable alternative to Annealed Langevin Sampling, and 2) a hybrid training formulation, composed of both Denoising Score Matching and adversarial objectives. By combining these two techniques and exploring different network architectures, we elevate score matching methods and obtain results competitive with state-of-the-art image generation on CIFAR-10.
△ Less
Submitted 10 October, 2020; v1 submitted 11 September, 2020;
originally announced September 2020.
-
DeepDrummer : Generating Drum Loops using Deep Learning and a Human in the Loop
Authors:
Guillaume Alain,
Maxime Chevalier-Boisvert,
Frederic Osterrath,
Remi Piche-Taillefer
Abstract:
DeepDrummer is a drum loop generation tool that uses active learning to learn the preferences (or current artistic intentions) of a human user from a small number of interactions. The principal goal of this tool is to enable an efficient exploration of new musical ideas. We train a deep neural network classifier on audio data and show how it can be used as the core component of a system that gener…
▽ More
DeepDrummer is a drum loop generation tool that uses active learning to learn the preferences (or current artistic intentions) of a human user from a small number of interactions. The principal goal of this tool is to enable an efficient exploration of new musical ideas. We train a deep neural network classifier on audio data and show how it can be used as the core component of a system that generates drum loops based on few prior beliefs as to how these loops should be structured.
We aim to build a system that can converge to meaningful results even with a limited number of interactions with the user. This property enables our method to be used from a cold start situation (no pre-existing dataset), or starting from a collection of audio samples provided by the user. In a proof of concept study with 25 participants, we empirically demonstrate that DeepDrummer is able to converge towards the preference of our subjects after a small number of interactions.
△ Less
Submitted 26 August, 2020; v1 submitted 10 August, 2020;
originally announced August 2020.
