NORDICS20 - Assets

Deep Learning on AWS

Amazon Web Services Resources EMEA

Issue link: https://emea-resources.awscloud.com/i/1242450

Contents of this Issue

Navigation

Page 5 of 50

Amazon Web Services Deep Learning on AWS Page 1 Overview The basic idea of deep learning has been around for decades. However, due to a recent surge in the digitization of information, organizations have amassed large amounts of data that are easily consumable by machine learning pipelines. Even more, our generation is spending more time on mobile phones and computers connected on social media and that has led to more data. Even in medicine, an x-ray image is stored as a digital record instead of a physical film. On one hand, the amount of data has exploded, but the performance of traditional machine learning algorithms such as logistic regression, decision trees, support vector machines, and others have plateaued even when fed with more data. In a turn of events, the small neural network has improved in accuracy of application; the medium neural network has improved in accuracy of application; and the deep neural network continues to improve in accuracy when fed with more data. We have yet to reach the limits of accuracy we can get by introducing more layers and more data in a deep neural network. Adding more layers to a neural network and providing more data helped improve the accuracy of deep learning applications. However, training the deep neural network was a hurdle because training requires access to powerful and often expensive compute infrastructure. Cloud computing solved this problem by offering on-demand GPUs in a cost effective and elastic manner, enabling large scale experimentation required to achieve the desired level of model accuracy. Although there are other tools under the broader umbrella of machine learning, such as probabilistic graph models, planning algorithms, search algorithms, and knowledge graphs, which are steadily improving, deep learning has improved exponentially and continues to break new ground. In this guide, we discuss the unique value proposition that Amazon Web Services (AWS) offers to support deep learning projects. You can leverage AWS innovation in the deep learning domain to improve the training time of deep learning jobs by using AWS optimized compute, storage, and network infrastructure. Deep learning jobs become more productive and agile from using AWS services and by offloading the undifferentiated heavy lifting involved in managing deep learning infrastructure and platform on AWS. AWS offers the deepest and broadest set of capabilities and flexibility that is required for the explorative nature of deep learning projects.

Articles in this issue

view archives of NORDICS20 - Assets - Deep Learning on AWS