Deep learning has led to tremendous advancements in the field of Artificial Intelligence. One caveat however is the substantial amount of compute needed to train these deep learning models. Training a benchmark dataset like ImageNet on a single machine with a modern GPU can take up to a week,distributing training on multiple machines has been observed to drastically bring this time down. Recent work has brought down ImageNet training time to a time as low as 4 minutes by using a cluster of 2048 GPUs. This paper surveys the various algorithms and techniques used to distribute training and presents the current state of the art for a modern distributed training framework. More specifically, we explore the synchronous and asynchronous variants of distributed Stochastic Gradient Descent, various All Reduce gradient aggregation strategies and best practices for obtaining higher throughout and lower latency over a cluster such as mixed precision training, large batch training and gradient compression. Read More
Tag Archives: Distributed Learning
Efficient Decentralized Deep Learning by Dynamic Model Averaging
We propose an efficient protocol for decentralized training of deep neural networks from distributed data sources. The proposed protocol allows to handle different phases of model training equally well and to quickly adapt to concept drifts. This leads to a reduction of communication by an order of magnitude compared to periodically communicating state-of-the-art approaches. Moreover, we derive a communication bound that scales well with the hardness of the serialized learning problem. The reduction in communication comes at almost no cost, as the predictive performance remains virtually unchanged. Indeed, the proposed protocol retains loss bounds of periodically averaging schemes. An extensive empirical evaluation validates major improvement of the trade-off between model performance and communication which could be beneficial for numerous decentralized learning applications, such as autonomous driving, or voice recognition and image classification on mobile phones. Read More
Incremental Convolutional Neural Network Training
Experimenting novel ideas on deep convolutional neural networks (DCNNs) with big datasets is hampered by the fact that network training requires huge computational resources in the terms of CPU and GPU power and hours. One option is to downscale the problem, e.g., less classes and less samples, but this is undesirable with DCNNs whose performance is largely data-dependent. In this work, we take an alternative route and downscale the networks and input images. For example, the ImageNet problem of 1,000 classes and 1,2M training images can be solved in hours on a commodity laptop without GPU by downscaling images and the network to the resolution of 8 8. We attempt to provide the solution to transfer the knowledge (parameters) of a trained DCNN with lower resolution to improve the efficiency of training a DCNN with higher resolution, and continue training incrementally until the full resolution is achieved. In our experiments, this approach achieves clear boost in computing time without the loss of performance. Read More
Parallel and Distributed Deep Learning
This report explores ways to parallelize/distribute deep learning in multi-core and distributed setting. We have analyzed (empirically) the speedup in training a CNN using conventional single core CPU and GPU and provide practical suggestions to improve training times. In the distributed setting, we study and analyze synchronous and asynchronous weight update algorithms (like Parallel SGD, ADMM and Downpour SGD) and come up with worst case asymptotic communication cost and computation time for each of the these algorithms. Read More
Demystifying Parallel and Distributed Deep Learning: An In-Depth Concurrency Analysis
Deep Neural Networks (DNNs) are becoming an important tool in modern computing applications. Accelerating their training is a major challenge and techniques range from distributed algorithms to low-level circuit design. In this survey, we describe the problem from a theoretical perspective, followed by approaches for its parallelization. We present trends in DNN architectures and the resulting implications on parallelization strategies. We then review and model the different types of concurrency in DNNs: from the single operator, through parallelism in network inference and training, to distributed deep learning. We discuss asynchronous stochastic optimization, distributed system architectures, communication schemes, and neural architecture search. Based on those approaches, we extrapolate potential directions for parallelism in deep learning. Read More
Distributed Deep Neural Networks over the Cloud, the Edge and End Devices
We propose distributed deep neural networks (DDNNs) over distributed computing hierarchies, consisting of the cloud, the edge (fog) and end devices. While being able to accommodate inference of a deep neural network (DNN) in the cloud, a DDNN also allows fast and localized inference using shallow portions of the neural network at the edge and end devices. When supported by a scalable distributed computing hierarchy, a DDNN can scale up in neural network size and scale out in geographical span. Due to its distributed nature, DDNNs enhance sensor fusion, system fault tolerance and data privacy for DNN applications. Read More
Large Scale Distributed Neural Network Training Through Online Distillation
Techniques such as ensembling and distillation promise model quality improvements when paired with almost any base model. However, due to increased testtime cost (for ensembles) and increased complexity of the training pipeline (for distillation), these techniques are challenging to use in industrial settings. In this paper we explore a variant of distillation which is relatively straightforward to use as it does not require a complicated multi-stage setup or many new hyperparameters. Our first claim is that online distillation enables us to use extra parallelism to fit very large datasets about twice as fast. Crucially, we can still speed up training even after we have already reached the point at which additional parallelism provides no benefit for synchronous or asynchronous stochastic gradient descent. Two neural networks trained on disjoint subsets of the data can share knowledge by encouraging each model to agree with the predictions the other model would have made. These predictions can come from a stale version of the other model so they can be safely computed using weights that only rarely get transmitted. Our second claim is that online distillation is a cost-effective way to make the exact predictions of a model dramatically more reproducible. We support our claims using experiments on the Criteo Display Ad Challenge dataset, ImageNet, and the largest to-date dataset used for neural language modeling, containing 6 × 1011 tokens and based on the Common Crawl repository of web data. Read More
Distributed Deep Learning, Part 1: An Introduction to Distributed Training of Neural Networks
Modern neurl network architectures trained on large data sets can obtain impressive performance across a wide variety of domains, from speech and image recognition, to natural language processing to industry-focused applications such as fraud detection and recommendation systems. But training these neural network models is computationally demanding. Although in recent years significant advances have been made in GPU hardware, network architectures and training methods, the fact remains that network training can take an impractically long time on a single machine. Fortunately, we are not restricted to a single machine: a significant amount of work and research has been conducted on enabling the efficient distributed training of neural networks. Read More
Communication-Efficient Learning of Deep Networks from Decentralized Data
Modern mobile devices have access to a wealth of data suitable for learning models, which in turn can greatly improve the user experience on the device. For example, language models can improve speech recognition and text entry, and image models can automatically select good photos. However, this rich data is often privacy sensitive, large in quantity, or both, which may preclude logging to the data center and training there using conventional approaches. We advocate an alternative that leaves the training data distributed on the mobile devices, and learns a shared model by aggregating locally-computed updates. We term this decentralized approach Federated Learning. Read More
Rick's Cafe AI 