REALIZE THE POTENTIAL OF MACHINE LEARNING
Systems that automatically learn without being programmed. It sounds so simple. But machine learning in real life — for organizations tasked with implementing the latest advancements or researchers and data scientists faced with finding value from lakes of data — requires a reliable partner who can provide the critical machine learning platforms and machine learning tools that are necessary to turn big data into action.
Cray is your machine learning partner.
We’ve been helping researchers and scientists solve the most complex problems in science and industry for over 40 years, and machine learning is no different.
Machine learning at supercomputer scale
Analytics allow you to understand the meaning behind your data. Machine learning and deep learning enable you to predict and act on that prediction with confidence. Underpinning both is the presence of big data that is captured, managed, optimized, and then used to train models that learn and can be used to make decisions.
Machine learning systems are beginning to transform industries: improving operational efficiency, creating new customer experiences, improving customer satisfaction, and increasing company value through competitive differentiation. Underlying this, IT infrastructure must address some constants:
- Never-ending accumulation and digitization of big data
- Steady growth of a machine learning portfolio as new use cases emerge
- Unyielding need for faster and more accurate results
We’ve taken some of the most widely used machine learning tools and platforms and made them available in our portfolio of supercomputer-powered systems.
Production-grade machine learning workflows – data preparation, model development, model training, model testing and inference/classification — are data-heavy and computationally intensive. The same supercomputer systems used to predict the weather, explore the cosmos, find and manage oil reserves, and make automobiles safer and quieter are ideally suited to power the discovery and automation being made possible by machine learning.
Machine learning platforms
Our Cray® Urika®-GX agile analytics platform is an integrated system designed for complex machine learning workflows:
- Data collection using tools like Apache® Kafka®
- Data cleansing, preparation and enrichment using data Apache® Spark™
- Machine learning using tools like MLlib and BigDL
- Advanced pattern matching and discovery using the Cray Graph Engine
Machine learning tools
Large-scale analytics, machine learning and deep learning have become a significant workload on HPC systems. The Urika®-XC analytics and artificial intelligence suite brings big data analytics, machine learning, distributed deep learning and large-scale graph processing to the massively scalable Cray® XC™ series supercomputer.
The Urika-XC suite brings together modern tools and languages to power analytics, machine learning and deep learning, including:
- The Apache Spark fast, general-purpose engine for large-scale data processing, with an extensive set of libraries and packages for analytics and machine learning uses (BigDL, Scala, Java, Python, R and MLlib)
- Distributed Dask and Python for open data science
- Jupyter Notebook for interactive data science
- Distributed deep learning with Google® TensorFlow™, augmented with the Cray Programming Environment Machine Learning Plugin, for unprecedented scale and ease of use