AWS re:Invent 2016

For many companies, recommendation systems solve important machine learning problems. But as recommendation systems grow to millions of users and millions of items, they pose significant challenges when deployed at scale. The user-item matrix can have trillions of entries (or more), most of which are zero. To make common ML techniques practical, sparse data requires special techniques. Learn how to use MXNet to build neural network models for recommendation systems that can scale efficiently to large sparse datasets.

In this session, we take a specific business problem—predicting Telco customer churn—and explore the practical aspects of building and evaluating an Amazon Machine Learning model. We explore considerations ranging from assigning a dollar value to applying the model using the relative cost of false positive and false negative errors. We discuss all aspects of putting Amazon ML to practical use, including how to build multiple models to choose from, put models into production, and update them. We also discuss using Amazon Redshift and Amazon S3 with Amazon ML.

Diabetic retinopathy (DR) is the leading cause of blindness for people aged 20 to 64, and afflicts more than 120 million people worldwide. Fortunately, vigilant monitoring greatly improves the chance to preserve one’s eyesight. This work used deep learning to analyze images of the retina and fundus for automated diagnosis of DR on a grading scale from 0 (normal) to 4 (severe). We achieved substantial improvement in accuracy compared to traditional approaches and continued advances by using a small auxiliary dataset that provided low-effort, high-value supervision. Data for training and testing, provided by the 2015 Kaggle Data Science Competition with over 80,000 high resolution images (>4 megapixels), required Amazon EC2 scalability to provide the GPU hardware needed to train a convolutional network with over 2 million parameters. For the competition, we focused on accurately modeling the scoring system, penalizing bad mistakes more severely, and combatting the over-prevalence of grade-0 examples in the dataset. We explored ideas first at low resolution on low-cost single-GPU instances. After finding the best methodology, we showed it could be scaled to equivalent improvements at high resolution, using the more expensive quad-GPU instances more effectively. This prototype model placed 15 out of 650 teams across the world with a kappa score of 0.78. We’ve now advanced the model via a new architecture that integrates the prototype and a new network specialized in finding dot hemorrhages, critical to identifying early DR. By annotating a small set of 200 images for hemorrhages, the performance jumped to a kappa of 0.82. We believe strategies that employ a bit more supervision for more effective learning are pivotal for cracking deep learning’s greatest weakness: its voracious appetite for data.

In this session, we show how to seamlessly transition VOD, live, and other advanced media workflows from on-premises deployments to the cloud. Cinépolis will provide an overview of their transcoding solution on AWS and how they have seamlessly expanded the solution increasing their customer reach. We'll show real world examples of the API calls used to configure and control all elements of the workflow including compression and origination. And how standard AWS services can be media-optimized with Elemental Technologies to form a robust live solution.

As Turner continues to make the transition from a traditional broadcast organization to a consumer-centric, data-driven media company, we are being challenged to re-think our approach to content supply. There is a need to achieve new levels of agility, flexibility and scalability to meet the rapidly evolving demands of our top media brands - including TBS, TNT, Cartoon Network, Adult Swim and CNN. To that end, we are transitioning the infrastructure that acquires, processes and distributes media for consumer-facing systems to the cloud. At the core of this environment is our Supply Chain Management application. The SCM app provides business and technical process management via an HTML based UI framework, State Machine, Rules Engine, Cost Model, Forms Service. We took advantage of several AWS specific services, including Lambda, S3, Dynamo DB, SNS, Elastic, Cloud Formation and Code Commit. The entire system is instance-less with all application code running in either the browser or within Lambda's. To ease development and debugging we created a method to run all JS libraries in the browser, switching to Lambda when we deploy with Code Commit. Cloud media processing infrastructure is BEING created on demand via an integration with SDVI. The SDVI and SCM apps exchange events and data via SNS and S3.

Media delivery requirements are continually changing, driven by accelerating mobile, tablet, smart TV, and set-top technology advances. Broadcasters need agile solutions to the changing media and entertainment landscape that don't require multiyear projects with large upfront investments. In this session, we walk through Discovery Communications' migration of its broadcast playout and channel origination to AWS. Discovery Communications is a leader in nonfiction media, reaching more than 3 billion cumulative viewers in 220 countries and territories. Traditionally, broadcast origination for content delivered to telecommunications companies, cable TV, and satellite has existed only in on-premises data centers. In this session, we walk through Discovery's migration of broadcast playout supporting hundreds of channels worldwide to AWS. We show how Discovery has not only reduced their TCO but also has improved their agility by launching new channels on demand. We also walk through how channel origination is being deployed in a secure, automated fashion, and with a level of high availability that exceeds what is possible in a traditional data center.

The process of making a film is highly complex, and comprises of multiple workflows across story development, pre-production, production, post-production and final distribution. Given the size and amount of media and assets associated with each stage, high performance infrastructure is often essential to meeting deadlines.

AWS provides a range of services and tools to help you create industry leading, cloud-enabled mobile apps that can securely scale to millions of users globally. Join Amit Patel, GM of AWS Mobile, to hear our vision for mobile apps and the cloud, industry trends, recent product launches, and success stories directly from our customers. We'll walk through and demo the AWS Mobile offerings for building compelling cloud-enabled mobile apps and for engaging your app users. You’ll learn how to use these offerings (serverless – API Gateway/Lambda, Cognito, and new services) to make it easy to develop both your iOS and Android frontend, as well as your mobile backend.

As GoPro expands into content networks and launches new products, new challenges have appeared. One of the most critical challenges facing GoPro during this period of rapid growth is their ability to make effective use of massive amounts of data. Every day, GoPro collects increasing amounts of data generated by internet connected consumer devices (smart cameras, smart drones), GoPro mobile apps, GoPro content networks, GoPro e-commerce sales, and social media. This data ranges from raw camera logs to refined and well-structured e-commerce datasets. In the past, it took GoPro months to understand new inbound data and determine how to transform or augment it for analysis. To streamline this process and bridge the gap between tech-savvy engineers and data-savvy analysts, GoPro is creating an analysis loop, which informs product usage trends and product insights. This analysis loop serves a large ecosystem of GoPro executives, product managers, engineers, data scientists, and business analysts through an integrated technology pipeline consisting of Apache Kafka, Apache Spark Streaming, Cloudera’s distribution of Hadoop, and Tableau’s Data Visualization Software as the end user analytical tool. Session sponsored by Tableau Software.

Companies around the world are using Amazon Simple Email Service (Amazon SES) to send millions of emails to their customers every day, and scaling linearly, at cost. In this session, you learn how to use the scalable and reliable infrastructure of Amazon SES. In addition, Netflix talks about their advanced Messaging program, their challenges, how SES helped them with their goals, and how they architected their solution for global scale and deliverability.