As computer architectures move towards multicore we must build a theoretical understanding of parallelism in machine learning. In this paper we focus on parallel inference in graphical models. We demonstrate that the natural, fully synchronous parallelization of belief propagation is highly inefficient. By bounding the achievable parallel performance in chain graphical models we develop a theoretical understanding of the parallel limitations of belief propagation. We then provide a new parallel belief propagation algorithm which achieves optimal performance. Using two challenging real-world tasks, we empirically evaluate the performance of our algorithm on large cyclic graphical models where we achieve near linear parallel scaling and out perform alternative algorithms. 1

We develop an online variational Bayes (VB) algorithm for Latent Dirichlet Allocation (LDA). Online LDA is based on online stochastic optimization with a natural gradient step, which we show converges to a local optimum of the VB objective function. It can handily analyze massive document collections, including those arriving in a stream. We study the performance of online LDA in several ways, including by fitting a 100-topic topic model to 3.3M articles from Wikipedia in a single pass. We demonstrate that online LDA finds topic models as good or better than those found with batch VB, and in a fraction of the time. 1

Latent Dirichlet analysis, or topic modeling, is a flexible latent variable framework for modeling high-dimensional sparse count data. Various learning algorithms have been developed in recent years, including collapsed Gibbs sampling, variational inference, and maximum a posteriori estimation, and this variety motivates the need for careful empirical comparisons. In this paper, we highlight the close connections between these approaches. We find that the main differences are attributable to the amount of smoothing applied to the counts. When the hyperparameters are optimized, the differences in performance among the algorithms diminish significantly. The ability of these algorithms to achieve solutions of comparable accuracy gives us the freedom to select computationally efficient approaches. Using the insights gained from this comparative study, we show how accurate topic models can be learned in several seconds on text corpora with thousands of documents. 1

Distributed learning is a problem of fundamental interest in machine learning and cognitive science. In this paper, we present asynchronous distributed learning algorithms for two well-known unsupervised learning frameworks: Latent Dirichlet Allocation (LDA) and Hierarchical Dirichlet Processes (HDP). In the proposed approach, the data are distributed across P processors, and processors independently perform Gibbs sampling on their local data and communicate their information in a local asynchronous manner with other processors. We demonstrate that our asynchronous algorithms are able to learn global topic models that are statistically as accurate as those learned by the standard LDA and HDP samplers, but with significant improvements in computation time and memory. We show speedup results on a 730-million-word text corpus using 32 processors, and we provide perplexity results for up to 1500 virtual processors. As a stepping stone in the development of asynchronous HDP, a parallel HDP sampler is also introduced. 1

personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission.

Topic models provide a powerful tool for analyzing large text collections by representing high dimensional data in a low dimensional subspace. Fitting a topic model given a set of training documents requires approximate inference techniques that are computationally expensive. With today’s large-scale, constantly expanding document collections, it is useful to be able to infer topic distributions for new documents without retraining the model. In this paper, we empirically evaluate the performance of several methods for topic inference in previously unseen documents, including methods based on Gibbs sampling, variational inference, and a new method inspired by text classification. The classificationbased inference method produces results similar to iterative inference methods, but requires only a single matrix multiplication. In addition to these inference methods, we present SparseLDA, an algorithm and data structure for evaluating Gibbs sampling distributions. Empirical results indicate that SparseLDA can be approximately 20 times faster than traditional LDA and provide twice the speedup of previously published fast sampling methods, while also using substantially less memory.

The recent emergence of Graphics Processing Units (GPUs) as general-purpose parallel computing devices provides us with new opportunities to develop scalable learning methods for massive data. In this work, we consider the problem of parallelizing two inference methods on GPUs for latent Dirichlet Allocation (LDA) models, collapsed Gibbs sampling (CGS) and collapsed variational Bayesian (CVB). To address limited memory constraints on GPUs, we propose a novel data partitioning scheme that effectively reduces the memory cost. This partitioning scheme also balances the computational cost on each multiprocessor and enables us to easily avoid memory access conflicts. We use data streaming to handle extremely large datasets. Extensive experiments showed that our parallel inference methods consistently produced LDA models with the same predictive power as sequential training methods did but with 26x speedup for CGS and 196x speedup for CVB on a GPU with 30 multiprocessors. The proposed partitioning scheme and data streaming make our approach scalable with more multiprocessors. Furthermore, they can be used as general techniques to parallelize other machine learning models. 1

Abstract. This paper presents PLDA, our parallel implementation of Latent Dirichlet Allocation on MPI and MapReduce. PLDA smooths out storage and computation bottlenecks and provides fault recovery for lengthy distributed computations. We show that PLDA can be applied to large, real-world applications and achieves good scalability. We have released MPI-PLDA to open source at http://code.google.com/p/plda under the Apache License. 1

Rapid growth in the amount of data available on social networking sites has made information retrieval increasingly challenging for users. In this paper, we propose a collaborative filtering method, Combinational Collaborative Filtering (CCF), to perform personalized community recommendations by considering multiple types of co-occurrences in social data at the same time. This filtering method fuses semantic and user information, then applies a hybrid training strategy that combines Gibbs sampling and Expectation-Maximization algorithm. To handle the large-scale dataset, parallel computing is used to speed up the model training. Through an empirical study on the Orkut dataset, we show CCF to be both effective and scalable.

As computer clusters become more common and the size of the problems encountered in the field of AI grows, there is an increasing demand for efficient parallel inference algorithms. We consider the problem of parallel inference on large factor graphs in the distributed memory setting of computer clusters. We develop a new efficient parallel inference algorithm, DBRSplash, which incorporates over-segmented graph partitioning, belief residual scheduling, and uniform work Splash operations. We empirically evaluate the DBRSplash algorithm on a 120 processor cluster and demonstrate linear to super-linear performance gains on large factor graph models. 1