Results 1  10
of
23
Large Scale Distributed Deep Networks
"... Recent work in unsupervised feature learning and deep learning has shown that being able to train large models can dramatically improve performance. In this paper, we consider the problem of training a deep network with billions of parameters using tens of thousands of CPU cores. We have developed a ..."
Abstract

Cited by 19 (6 self)
 Add to MetaCart
(Show Context)
Recent work in unsupervised feature learning and deep learning has shown that being able to train large models can dramatically improve performance. In this paper, we consider the problem of training a deep network with billions of parameters using tens of thousands of CPU cores. We have developed a software framework called DistBelief that can utilize computing clusters with thousands of machines to train large models. Within this framework, we have developed two algorithms for largescale distributed training: (i) Downpour SGD, an asynchronous stochastic gradient descent procedure supporting a large number of model replicas, and (ii) Sandblaster, a framework that supports a variety of distributed batch optimization procedures, including a distributed implementation of LBFGS. Downpour SGD and Sandblaster LBFGS both increase the scale and speed of deep network training. We have successfully used our system to train a deep network 30x larger than previously reported in the literature, and achieves stateoftheart performance on ImageNet, a visual object recognition task with 16 million images and 21k categories. We show that these same techniques dramatically accelerate the training of a more modestly sized deep network for a commercial speech recognition service. Although we focus on and report performance of these methods as applied to training large neural networks, the underlying algorithms are applicable to any gradientbased machine learning algorithm. 1
A reliable effective terascale linear learning system
, 2011
"... We present a system and a set of techniques for learning linear predictors with convex losses on terascale data sets, with trillions of features,1 billions of training examples and millions of parameters in an hour using a cluster of 1000 machines. Individually none of the component techniques are n ..."
Abstract

Cited by 19 (2 self)
 Add to MetaCart
(Show Context)
We present a system and a set of techniques for learning linear predictors with convex losses on terascale data sets, with trillions of features,1 billions of training examples and millions of parameters in an hour using a cluster of 1000 machines. Individually none of the component techniques are new, but the careful synthesis required to obtain an efficient implementation is. The result is, up to our knowledge, the most scalable and efficient linear learning system reported in the literature.2 We describe and thoroughly evaluate the components of the system, showing the importance of the various design choices.
Recent Advances of Largescale Linear Classification
"... Linear classification is a useful tool in machine learning and data mining. For some data in a rich dimensional space, the performance (i.e., testing accuracy) of linear classifiers has shown to be close to that of nonlinear classifiers such as kernel methods, but training and testing speed is much ..."
Abstract

Cited by 8 (3 self)
 Add to MetaCart
(Show Context)
Linear classification is a useful tool in machine learning and data mining. For some data in a rich dimensional space, the performance (i.e., testing accuracy) of linear classifiers has shown to be close to that of nonlinear classifiers such as kernel methods, but training and testing speed is much faster. Recently, many research works have developed efficient optimization methods to construct linear classifiers and applied them to some largescale applications. In this paper, we give a comprehensive survey on the recent development of this active research area.
Better MiniBatch Algorithms via Accelerated Gradient Methods
"... Minibatch algorithms have been proposed as a way to speedup stochastic convex optimization problems. We study how such algorithms can be improved using accelerated gradient methods. We provide a novel analysis, which shows how standard gradient methods may sometimes be insufficient to obtain a sig ..."
Abstract

Cited by 6 (3 self)
 Add to MetaCart
(Show Context)
Minibatch algorithms have been proposed as a way to speedup stochastic convex optimization problems. We study how such algorithms can be improved using accelerated gradient methods. We provide a novel analysis, which shows how standard gradient methods may sometimes be insufficient to obtain a significant speedup and propose a novel accelerated gradient algorithm, which deals with this deficiency, enjoys a uniformly superior guarantee and works well in practice. 1
Structure and Dynamics of Information Pathways in Online Media
"... Diffusion of information, spread of rumors and infectious diseases are all instances of stochastic processes that occur over the edges of an underlying network. Many times networks over which contagions spread are unobserved, and such networks are often dynamic and change over time. In this paper, w ..."
Abstract

Cited by 4 (1 self)
 Add to MetaCart
(Show Context)
Diffusion of information, spread of rumors and infectious diseases are all instances of stochastic processes that occur over the edges of an underlying network. Many times networks over which contagions spread are unobserved, and such networks are often dynamic and change over time. In this paper, we investigate the problem of inferring dynamic networks based on information diffusion data. We assume there is an unobserved dynamic network that changes over time, while we observe the results of a dynamic process spreading over the edges of the network. The task then is to infer the edges and the dynamics of the underlying network. We develop an online algorithm that relies on stochastic convex optimization to efficiently solve the dynamic network inference problem. We apply our algorithm to information diffusion among 3.3 million mainstream media and blog sites and experiment with more than 179 million different pieces of information spreading over the network in a one year period. We study the evolution of information pathways in the online media space and find interesting insights. Information pathways for general recurrent topics are more stable across time than for ongoing news events. Clusters of news media sites and blogs often emerge and vanish in matter of days for ongoing news events. Major social movements and events involving civil population, such as the Libyan’s civil war or Syria’s uprise, lead to an increased amount of information pathways among blogs as well as in the overall increase in the network centrality of blogs and social media sites.
MLbase: A Distributed Machinelearning System
"... Machine learning (ML) and statistical techniques are key to transforming big data into actionable knowledge. In spite of the modern primacy of data, the complexity of existing ML algorithms is often overwhelming—many users do not understand the tradeoffs and challenges of parameterizing and choosin ..."
Abstract

Cited by 3 (0 self)
 Add to MetaCart
(Show Context)
Machine learning (ML) and statistical techniques are key to transforming big data into actionable knowledge. In spite of the modern primacy of data, the complexity of existing ML algorithms is often overwhelming—many users do not understand the tradeoffs and challenges of parameterizing and choosing between different learning techniques. Furthermore, existing scalable systems that support machine learning are typically not accessible to ML researchers without a strong background in distributed systems and lowlevel primitives. In this work, we present our vision for MLbase, a novel system harnessing the power of machine learning for both endusers and ML researchers. MLbase provides (1) a simple declarative way to specify ML tasks, (2) a novel optimizer to select and dynamically adapt the choice of learning algorithm, (3) a set of highlevel operators to enable ML researchers to scalably implement a wide range of ML methods without deep systems knowledge, and (4) a new runtime optimized for the dataaccess patterns of these highlevel operators. 1.
Minimizing Finite Sums with the Stochastic Average Gradient
, 2013
"... We propose the stochastic average gradient (SAG) method for optimizing the sum of a finite number of smooth convex functions. Like stochastic gradient (SG) methods, the SAG method’s iteration cost is independent of the number of terms in the sum. However, by incorporating a memory of previous gradie ..."
Abstract

Cited by 3 (0 self)
 Add to MetaCart
We propose the stochastic average gradient (SAG) method for optimizing the sum of a finite number of smooth convex functions. Like stochastic gradient (SG) methods, the SAG method’s iteration cost is independent of the number of terms in the sum. However, by incorporating a memory of previous gradient values the SAG method achieves a faster convergence rate than blackbox SG methods. The convergence rate is improved from O(1 / √ k) to O(1/k) in general, and when the sum is stronglyconvex the convergence rate is improved from the sublinear O(1/k) to a linear convergence rate of the form O(ρ k) for ρ < 1. Further, in many cases the convergence rate of the new method is also faster than blackbox deterministic gradient methods, in terms of the number of gradient evaluations. Numerical experiments indicate that the new algorithm often dramatically outperforms existing SG and deterministic gradient methods, and that the performance may be further improved through the use of nonuniform sampling strategies. 1
Minibatch primal and dual methods for SVMs
 In 30th International Conference on Machine Learning
, 2013
"... We address the issue of using minibatches in stochastic optimization of SVMs. We show that the same quantity, the spectral norm of the data, controls the parallelization speedup obtained for both primal stochastic subgradient descent (SGD) and stochastic dual coordinate ascent (SCDA) methods and us ..."
Abstract

Cited by 3 (2 self)
 Add to MetaCart
(Show Context)
We address the issue of using minibatches in stochastic optimization of SVMs. We show that the same quantity, the spectral norm of the data, controls the parallelization speedup obtained for both primal stochastic subgradient descent (SGD) and stochastic dual coordinate ascent (SCDA) methods and use it to derive novel variants of minibatched SDCA. Our guarantees for both methods are expressed in terms of the original nonsmooth primal problem based on the hingeloss. 1.
Scalable Coordinate Descent Approaches to Parallel Matrix Factorization for Recommender Systems
"... Abstract—Matrix factorization, when the matrix has missing values, has become one of the leading techniques for recommender systems. To handle webscale datasets with millions of users and billions of ratings, scalability becomes an important issue. Alternating Least Squares (ALS) and Stochastic Gra ..."
Abstract

Cited by 3 (0 self)
 Add to MetaCart
(Show Context)
Abstract—Matrix factorization, when the matrix has missing values, has become one of the leading techniques for recommender systems. To handle webscale datasets with millions of users and billions of ratings, scalability becomes an important issue. Alternating Least Squares (ALS) and Stochastic Gradient Descent (SGD) are two popular approaches to compute matrix factorization. There has been a recent flurry of activity to parallelize these algorithms. However, due to the cubic time complexity in the target rank, ALS is not scalable to largescale datasets. On the other hand, SGD conducts efficient updates but usually suffers from slow convergence that is sensitive to the parameters. Coordinate descent, a classical optimization approach, has been used for many other largescale problems, but its application to matrix factorization for recommender systems has not been explored thoroughly. In this paper, we show that coordinate descent based methods have a more efficient update rule compared to ALS, and are faster and have more stable convergence than SGD. We study different update sequences and propose the CCD++ algorithm, which updates rankone factors one by one. In addition, CCD++ can be easily parallelized on both multicore and distributed systems. We empirically show that CCD++ is much faster than ALS and SGD in both settings. As an example, on a synthetic dataset with 2 billion ratings, CCD++ is 4 times faster than both SGD and ALS using a distributed system with 20 machines. KeywordsRecommender systems, Matrix factorization, Low rank approximation, Parallelization.
Online Learning under Delayed Feedback
"... Online learning with delayed feedback has received increasing attention recently due to its several applications in distributed, webbased learning problems. In this paper we provide a systematic study of the topic, and analyze the effect of delay on the regret of online learning algorithms. Somewha ..."
Abstract

Cited by 2 (0 self)
 Add to MetaCart
(Show Context)
Online learning with delayed feedback has received increasing attention recently due to its several applications in distributed, webbased learning problems. In this paper we provide a systematic study of the topic, and analyze the effect of delay on the regret of online learning algorithms. Somewhat surprisingly, it turns out that delay increases the regret in a multiplicative way in adversarial problems, and in an additive way in stochastic problems. We give metaalgorithms that transform, in a blackbox fashion, algorithms developed for the nondelayed case into ones that can handle the presence of delays in the feedback loop. Modifications of the wellknown UCB algorithm are also developed for the bandit problem with delayed feedback, with the advantage over the metaalgorithms that they can be implemented with lower complexity. 1.