Subscribe to the gold package and get unlimited access to Shamra Academy
Register a new userStochastic variational inference for large-scale discrete choice models using adaptive batch sizes
87
0
0.0
(
0
)
Added by
Linda S. L. Tan
Publication date
2014
fields
Mathematical Statistics
and research's language is
English
Authors
Linda S. L. Tan
Ask ChatGPT about the research
No Arabic abstract
Discrete choice models describe the choices made by decision makers among alternatives and play an important role in transportation planning, marketing research and other applications. The mixed multinomial logit (MMNL) model is a popular discrete choice model that captures heterogeneity in the preferences of decision makers through random coefficients. While Markov chain Monte Carlo methods provide the Bayesian analogue to classical procedures for estimating MMNL models, computations can be prohibitively expensive for large datasets. Approximate inference can be obtained using variational methods at a lower computational cost with competitive accuracy. In this paper, we develop variational methods for estimating MMNL models that allow random coefficients to be correlated in the posterior and can be extended easily to large-scale datasets. We explore three alternatives: (1) Laplace variational inference, (2) nonconjugate variational message passing and (3) stochastic linear regression. Their performances are compared using real and simulated data. To accelerate convergence for large datasets, we develop stochastic variational inference for MMNL models using each of the above alternatives. Stochastic variational inference allows data to be processed in minibatches by optimizing global variational parameters using stochastic gradient approximation. A novel strategy for increasing minibatch sizes adaptively within stochastic variational inference is proposed.
rate research
Read More
Variational Inference makes a trade-off between the capacity of the variational family and the tractability of finding an approximate posterior distribution. Instead, Boosting Variational Inference allows practitioners to obtain increasingly good posterior approximations by spending more compute. The main obstacle to widespread adoption of Boosting Variational Inference is the amount of resources necessary to improve over a strong Variational Inference baseline. In our work, we trace this limitation back to the global curvature of the KL-divergence. We characterize how the global curvature impacts time and memory consumption, address the problem with the notion of local curvature, and provide a novel approximate backtracking algorithm for estimating local curvature. We give new theoretical convergence rates for our algorithms and provide experimental validation on synthetic and real-world datasets.
Stochastic variational inference for collapsed models has recently been successfully applied to large scale topic modelling. In this paper, we propose a stochastic collapsed variational inference algorithm for hidden Markov models, in a sequential data setting. Given a collapsed hidden Markov Model, we break its long Markov chain into a set of short subchains. We propose a novel sum-product algorithm to update the posteriors of the subchains, taking into account their boundary transitions due to the sequential dependencies. Our experiments on two discrete datasets show that our collapsed algorithm is scalable to very large datasets, memory efficient and significantly more accurate than the existing uncollapsed algorithm.
Human decision making underlies data generating process in multiple application areas, and models explaining and predicting choices made by individuals are in high demand. Discrete choice models are widely studied in economics and computational social sciences. As digital social networking facilitates information flow and spread of influence between individuals, new advances in modeling are needed to incorporate social information into these models in addition to characteristic features affecting individual choices. In this paper, we propose two novel models with scalable training algorithms: local logistics graph regularization (LLGR) and latent class graph regularization (LCGR) models. We add social regularization to represent similarity between friends, and we introduce latent classes to account for possible preference discrepancies between different social groups. Training of the LLGR model is performed using alternating direction method of multipliers (ADMM), and training of the LCGR model is performed using a specialized Monte Carlo expectation maximization (MCEM) algorithm. Scalability to large graphs is achieved by parallelizing computation in both the expectation and the maximization steps. The LCGR model is the first latent class classification model that incorporates social relationships among individuals represented by a given graph. To evaluate our two models, we consider three classes of data to illustrate a typical large-scale use case in internet and social media applications. We experiment on synthetic datasets to empirically explain when the proposed model is better than vanilla classification models that do not exploit graph structure. We also experiment on real-world data, including both small scale and large scale real-world datasets, to demonstrate on which types of datasets our model can be expected to outperform state-of-the-art models.
Gradient-based approximate inference methods, such as Stein variational gradient descent (SVGD), provide simple and general-purpose inference engines for differentiable continuous distributions. However, existing forms of SVGD cannot be directly applied to discrete distributions. In this work, we fill this gap by proposing a simple yet general framework that transforms discrete distributions to equivalent piecewise continuous distributions, on which the gradient-free SVGD is applied to perform efficient approximate inference. The empirical results show that our method outperforms traditional algorithms such as Gibbs sampling and discontinuous Hamiltonian Monte Carlo on various challenging benchmarks of discrete graphical models. We demonstrate that our method provides a promising tool for learning ensembles of binarized neural network (BNN), outperforming other widely used ensemble methods on learning binarized AlexNet on CIFAR-10 dataset. In addition, such transform can be straightforwardly employed in gradient-free kernelized Stein discrepancy to perform goodness-of-fit (GOF) test on discrete distributions. Our proposed method outperforms existing GOF test methods for intractable discrete distributions.
For large scale on-line inference problems the update strategy is critical for performance. We derive an adaptive scan Gibbs sampler that optimizes the update frequency by selecting an optimum mini-batch size. We demonstrate performance of our adaptive batch-size Gibbs sampler by comparing it against the collapsed Gibbs sampler for Bayesian Lasso, Dirichlet Process Mixture Models (DPMM) and Latent Dirichlet Allocation (LDA) graphical models.
Log in to be able to interact and post comments
comments
Fetching comments
Sign in to be able to follow your search criteria
