Stein Self-Repulsive Dynamics: Benefits From Past Samples
arXiv:2002.09070
Abstract
We propose a new Stein self-repulsive dynamics for obtaining diversified samples from intractable un-normalized distributions. Our idea is to introduce Stein variational gradient as a repulsive force to push the samples of Langevin dynamics away from the past trajectories. This simple idea allows us to significantly decrease the auto-correlation in Langevin dynamics and hence increase the effective sample size. Importantly, as we establish in our theoretical analysis, the asymptotic stationary distribution remains correct even with the addition of the repulsive force, thanks to the special properties of the Stein variational gradient. We perform extensive empirical studies of our new algorithm, showing that our method yields much higher sample efficiency and better uncertainty estimation than vanilla Langevin dynamics.
References in corpus (15)
- Escaping free-energy minima
- MCMC using Hamiltonian dynamics
- Weight Uncertainty in Neural Networks
- Basic Properties of Strong Mixing Conditions. A Survey and Some Open Questions
- Reinforcement Learning with Deep Energy-Based Policies
- Stein Variational Gradient Descent as Gradient Flow
- Stein Variational Policy Gradient
- Stein Variational Gradient Descent With Matrix-Valued Kernels
- Understanding and Accelerating Particle-Based Variational Inference
- Stein Variational Gradient Descent as Moment Matching
- Stein Variational Gradient Descent Without Gradient
- Stochastic Gradient MCMC with Repulsive Forces
- Stein Variational Adaptive Importance Sampling
- Non-convex Learning via Replica Exchange Stochastic Gradient MCMC
- A Contour Stochastic Gradient Langevin Dynamics Algorithm for Simulations of Multi-modal Distributions