More Like this

1. Decentralized Stochastic Gradient Langevin Dynamics and Hamiltonian Monte Carlo

   Gürbüzbalaban, M.; Gao, X.; Hu, Y.; Zhu, L. (January 2021, Journal of machine learning research)

   Stochastic gradient Langevin dynamics (SGLD) and stochastic gradient Hamiltonian Monte Carlo (SGHMC) are two popular Markov Chain Monte Carlo (MCMC) algorithms for Bayesian inference that can scale to large datasets, allowing to sample from the posterior distribution of the parameters of a statistical model given the input data and the prior distribution over the model parameters. However, these algorithms do not apply to the decentralized learning setting, when a network of agents are working collaboratively to learn the parameters of a statistical model without sharing their individual data due to privacy reasons or communication constraints. We study two algorithms: Decentralized SGLD (DE-SGLD) and Decentralized SGHMC (DE-SGHMC) which are adaptations of SGLD and SGHMC methods that allow scaleable Bayesian inference in the decentralized setting for large datasets. We show that when the posterior distribution is strongly log-concave and smooth, the iterates of these algorithms converge linearly to a neighborhood of the target distribution in the 2-Wasserstein distance if their parameters are selected appropriately. We illustrate the efficiency of our algorithms on decentralized Bayesian linear regression and Bayesian logistic regression problems 

   more » « less
2. Projected Stochastic Gradient Langevin Algorithms for Constrained Sampling and Non-Convex Learning

   Lamperski, A. (August 2021, Proceedings of Thirty Fourth Conference on Learning Theory)

   Belkin, M.; Kpotufe, S. (Ed.)

   Langevin algorithms are gradient descent methods with additive noise. They have been used for decades in Markov Chain Monte Carlo (MCMC) sampling, optimization, and learning. Their convergence properties for unconstrained non-convex optimization and learning problems have been studied widely in the last few years. Other work has examined projected Langevin algorithms for sampling from log-concave distributions restricted to convex compact sets. For learning and optimization, log-concave distributions correspond to convex losses. In this paper, we analyze the case of non-convex losses with compact convex constraint sets and IID external data variables. We term the resulting method the projected stochastic gradient Langevin algorithm (PSGLA). We show the algorithm achieves a deviation of 𝑂(𝑇−1/4(𝑙𝑜𝑔𝑇)1/2) from its target distribution in 1-Wasserstein distance. For optimization and learning, we show that the algorithm achieves 𝜖-suboptimal solutions, on average, provided that it is run for a time that is polynomial in 𝜖 and slightly super-exponential in the problem dimension. 

   more » « less
3. Adaptive Quantum Simulated Annealing for Bayesian Inference and Estimating Partition Functions

   Harrow, Aram W.; Wei, AY (January 2020, Proceedings of the 2020 ACM-SIAM Symposium on Discrete Algorithms)

   Markov chain Monte Carlo algorithms have important applications in counting problems and in machine learning problems, settings that involve estimating quantities that are difficult to compute exactly. How much can quantum computers speed up classical Markov chain algorithms? In this work we consider the problem of speeding up simulated annealing algorithms, where the stationary distributions of the Markov chains are Gibbs distributions at temperatures specified according to an annealing schedule. We construct a quantum algorithm that both adaptively constructs an annealing schedule and quantum samples at each temperature. Our adaptive annealing schedule roughly matches the length of the best classical adaptive annealing schedules and improves on nonadaptive temperature schedules by roughly a quadratic factor. Our dependence on the Markov chain gap matches other quantum algorithms and is quadratically better than what classical Markov chains achieve. Our algorithm is the first to combine both of these quadratic improvements. Like other quantum walk algorithms, it also improves on classical algorithms by producing “qsamples” instead of classical samples. This means preparing quantum states whose amplitudes are the square roots of the target probability distribution. In constructing the annealing schedule we make use of amplitude estimation, and we introduce a method for making amplitude estimation nondestructive at almost no additional cost, a result that may have independent interest. Finally we demonstrate how this quantum simulated annealing algorithm can be applied to the problems of estimating partition functions and Bayesian inference. 

   more » « less
4. High-dimensional nonlinear Bayesian inference of poroelastic fields from pressure data

   https://doi.org/10.1177/10812865221140840  

   Karimi, Mina; Massoudi, Mehrdad; Dayal, Kaushik; Pozzi, Matteo (February 2023, Mathematics and Mechanics of Solids)

   We investigate solution methods for large-scale inverse problems governed by partial differential equations (PDEs) via Bayesian inference. The Bayesian framework provides a statistical setting to infer uncertain parameters from noisy measurements. To quantify posterior uncertainty, we adopt Markov Chain Monte Carlo (MCMC) approaches for generating samples. To increase the efficiency of these approaches in high-dimension, we make use of local information about gradient and Hessian of the target potential, also via Hamiltonian Monte Carlo (HMC). Our target application is inferring the field of soil permeability processing observations of pore pressure, using a nonlinear PDE poromechanics model for predicting pressure from permeability. We compare the performance of different sampling approaches in this and other settings. We also investigate the effect of dimensionality and non-gaussianity of distributions on the performance of different sampling methods. 

   more » « less
5. Structured Stochastic Gradient MCMC

   Alexos, Antonios; Boyd, Alex; Mandt, Stephan (January 2022, Proceedings of Machine Learning Research)

   Stochastic gradient Markov Chain Monte Carlo (SGMCMC) is a scalable algorithm for asymptotically exact Bayesian inference in parameter-rich models, such as Bayesian neural networks. However, since mixing can be slow in high dimensions, practitioners often resort to variational inference (VI). Unfortunately, VI makes strong assumptions on both the factorization and functional form of the posterior. To relax these assumptions, this work proposes a new non-parametric variational inference scheme that combines ideas from both SGMCMC and coordinate-ascent VI. The approach relies on a new Langevin-type algorithm that operates on a "self-averaged" posterior energy function, where parts of the latent variables are averaged over samples from earlier iterations of the Markov chain. This way, statistical dependencies between coordinates can be broken in a controlled way, allowing the chain to mix faster. This scheme can be further modified in a "dropout" manner, leading to even more scalability. We test our scheme for ResNet-20 on CIFAR-10, SVHN, and FMNIST. In all cases, we find improvements in convergence speed and/or final accuracy compared to SGMCMC and parametric VI. 

   more » « less