NVIDIA GPU-Accelerated Analytics Roadshow Coming to EuropeJanuary 24, 2018
Fueled by massive data. Driven by deep insights. Powered by lightning-fast GPUs.
With the explosion of AI, traditional enterprise analytics solutions are no longer sufficient to handle massive amounts of data fast. Instead, GPU-accelerated analytics and deep learning applications have become crucial to success.
Businesses need to build a new set of muscles to be more productive, effective, insightful and automated. They need to adapt as AI and accelerated analytics technologies make their way into every facet of modern business, from fraud detection to cybersecurity to consumer experience.
Join us in Munich (Jan. 30), London (Feb. 6) and Paris (Feb. 8) to learn more about how NVIDIA and our GPU-accelerated partners are changing the traditional analytics ecosystem. These solutions work well for a variety of industries. Here are a few highlights:
H2O.ai: GPU-Powered Automatic Machine Learning for the Enterprise
H2O.ai Driverless AI, a GPU-accelerated machine learning platform, provides expert recipes, built by Kaggle Grandmasters, for data scientists and business users to train their machine learning models for faster and interpretable insights.
For instance it could take upto 30+ hours to build a well tuned GBM algorithm. The model preparation is complicated due to huge volume and various forms of data. The constant model re-training and re-evaluation are not avoidable because of the ever-changing nature of customer data and the need to improve model prediction accuracy. This doesn’t even include all the feature engineering effort required to get the most accuracy.
On top of that, there is a huge shortage of expert data science talent. The seemingly endless wait for dedicated resources means further delays in gaining business insights. And finally, interpreting the results of complex black box models is extremely difficult and a barrier toward enterprise adoption of AI.
H2O.ai is solving the problems of slow processing, model complexity and lack of expertise.
In our European roadshow, H2O.ai will cover how finance customer Paypal is using Driverless AI for fraud detection to prevent collusion and other organized crimes. Paypal leveraged one year of training data — which is 1.5 billion edges and 0.5 million nodes — trained on Driverless AI, powered by four NVIDIA Tesla P100 GPU accelerators, resulting in 6x faster results than on CPUs.
Kinetica: GPU-Accelerated Database
Kinetica is a distributed, in-memory, GPU-accelerated database. It provides real-time analytics on data in motion and at rest with at least 10-100x faster performance at 1/10 the cost of traditional systems.
Digital businesses face technology challenges as they try to cope with growing data volumes and data speed, ever more complex analysis of that data and making real-time decisions based on deep insights.
While relational database management systems and data warehouse technologies can store growing volumes of data, they struggle with analyzing this data for real-time decision making and also come with a very large price tag.
Even though storing large amounts of data becomes more affordable with technologies like Hadoop, their complexity and batch-oriented approach significantly delays time to insights.
And analyzing data, particularly perishable data at the same time as it’s ingested, needs next-generation technology. Legacy data systems or Hadoop infrastructures can’t keep up with streaming data analysis.
In our roadshow, Kinetica will share an example of how in the cybersecurity space, enterprises need a secure and resilient system to support high-density workloads and mitigate threats from all stages. This customer’s main objective is to identify a broad range of malicious activities in real time with confidence to protect their users against current and emerging cyber threats.
With a deep collection of all network traffic data, infrastructure data and log data, Kinetica helps the customer provide additional device context and insights into user activities. Additionally, Kinetica’s advanced in-database analytics via user-defined functions (UDFs) allows data scientists to operationalize machine learning libraries such as TensorFlow, Caffe and Torch (or custom logic) — bridging the gap between data science, data analytics and business intelligence, and gaining deeper insights.
MapD: GPU-Powered SQL Engine and Visualization System
MapD is an open source GPU-accelerated SQL engine and visualization system, enabling the exploration of billions of rows of data with millisecond latency at petabyte scale. In our roadshow, MapD will provide insights into buying patterns for different brands and products by combining granular consumer demographics data with credit-card purchasing data. These insights could ultimately be used to drive trading strategies for associated securities.
The first dataset contains detailed demographic data for the nearly 40 million zip+4 districts, courtesy of data provider Axciom. MapD explores this data with its geospatial rendering capabilities. Then, using another dataset of approximately 500 million credit card transactions, courtesy of Quandl, MapD shows how quickly you can drill into consumer buying patterns, such as revenue estimates for each iPhone launch.
Even greater insight comes from joining the two datasets. By creating an on-the-fly join between the geospatial coordinates of the credit card purchasing data and the zip+4 demographic data, MapD can identify correlations between things like income, wealth, education and political persuasion on purchasing decisions.
For example, it becomes clear that Netflix gets more revenue per capita from the lower middle and upper classes, while Home Depot is popular in Republican areas. And Uber seems to have started as a luxury of the rich, but has begun percolating down to the middle and upper middle classes. Investors could presumably use such insights to size future market opportunities for a consumer brand, in turn helping them develop trading strategies.
Last but not least, MapD will conclude the demo by using the GPU Data Frame/Apache Arrow enabled pymapd to quickly pull data out of MapD and run regressions to pinpoint the demographic factors that drive spend for various brands.
If you’re in Europe, come join us and watch the live demos to learn more about how GPU-accelerated analytics and machine learning solutions can help accelerate your enterprise workloads.