IBM Machine Learning leverages parts of the Watson supercomputer to be used with the IBM z System Mainframe. The tech giant's initiative aims to train and deploy analytics models in the private cloud.
IBM Cloud-Based Machine Learning
According to Tech Republic, IBM plans to bring soon some of the core machine learning technology from IBM Watson to the private cloud and mainframes, as the company announced on Wednesday, Feb. 15. The tech giant has called its new cognitive platform IBM Machine Learning. The upcoming machine learning platform will be launched first on the z System mainframe.
IBM described IBM Machine Learning in a press release as a platform that will facilitate creating, training and deploying high volume of analytic models in the private cloud. The new service could help enterprise data scientists achieve valuable insights more quickly. IBM Machine Learning is leveraging core Watson technologies in order to accelerate the adoption of machine learning, according to the statement of Rob Thomas, general manager for IBM Analytics, published in the press release.
IBM Machine Learning's Features
Every day, the z System mainframe handles billions of transactions from organizations in insurance, banking, retail, government, and more, as IBM pointed out in the release. For instance, the release said that healthcare could use machine learning service to better tailor offerings to patients and retail could use it to examine the day's trends in real-time. According to the release, IBM Machine Learning can help create and train operational analytic models by using any popular machine learning framework, any language and any transactional data type.
According to InfoWorld, IBM Machine Learning for z/OS is a collection of popular frameworks, rather than a single machine learning framework. Languages such as TensorFlow, Apache SparkML and H2O are packaged with bindings to common languages used in the trade, like Scala, Java and Python. IBM is pushing the new machine learning service as a pipeline for using tools for each step of the process of building, running and managing machine learning models.
This kind of convenience was really needed. Even as the number of frameworks for machine learning is growing, in order to create end-to-end production pipelines for training and working with models developers still have to perform a lot of heavy labor. In time, the IBM Machine learning for z/OS could serve as a platform for a complete solution covering every phase of machine learning.
The new cloud-based machine learning platform will also make use of IBM Research's Cognitive Automation for Data Scientists to help data scientists select the best algorithm for their work. According to the release, IBM Machine Learning for z/OS could lead to more valuable insights into the huge amount of data processed by the IBM z Systems mainframe, reaching around 2.5 billion transactions daily. The risk and latency are considerable reduced, as the data remains on the system.
The new service will only be available on z/OS for now, but eventually will extend to IBM POWER systems in the near future. IBM also launched the Watson Discovery Service in late 2016, to make big data analytics easily accessible to companies that have limited access to data science resources. IBM is positioning itself with the launch of IBM Machine Learning as an enterprise provider than can fill in a company's big data gaps and machine learning, helping them to remain competitive without dedicated in-house talent.
