HPE has released a pair of systems designed to broaden adoption and accelerate deployment of machine learning in enterprises. Swarm Learning aims to bring the wisdom of crowds to machine learning modeling without sacrificing security, while the machine learning development system is intended to offer a unique training solution to companies that would otherwise have had to design and build their own machine learning infrastructure. .
The machine learning development system is available in several different sized physical footprints, but the company says a “small configuration” uses an Apollo 6500 Gen10 compute server to provide the power needed for learning training. machine, HPE ProLiant DL325 servers and Aruba CX 6300 switches for system component management and NVIDIA’s Quantum InfiniBand networking platform, as well as specialized machine learning development environment management software suites and HPE Performance Cluster.
New system brings HPC computing to machine learning
Essentially, this is about bringing HPC (high performance computing) capabilities to enterprise machine learning, which would typically require companies to design their own, according to IDC research vice president Peter Rutten. systems.
“It’s the kind of system companies are really looking for, now that AI is more mature,” he said. “The biggest hurdle to bringing AI into your business is that you have to build the system.” Using cloud resources might be an option for some businesses, but the data required for AI models tends to be sensitive and business-critical, so some businesses might be hesitant about this option even though the restrictions regulations in some industries make it downright impossible for others.
Swarm Learning decentralizes machine learning
The sensitive nature of machine learning data is the problem HPE is trying to solve with its other new product, Swarm Learning. It is a decentralized framework that uses containerization to achieve two goals: first, it allows machine learning to take place on edge systems, without the need for a round trip to a data center central, allowing businesses to get accurate information faster than they otherwise would. to be able to. Second, it allows peer companies to share AI model learning outcomes with each other, potentially creating industry-wide benefits without requiring companies to share the underlying data with each other.
“So if you have seven hospitals that are all trying to solve problems with AI model training, but they can’t share data, then you have limited AI training,” Rutten said. This yields low precision models with potential bias built in, depending on hospital patient demographics and a host of other factors. “In order to solve this problem… swarm learning does not share data, but it shares the results of model training at each location and combines them into a trained model.”
Rutten noted that swarm learning is a relatively new technique, which means widespread adoption may be slow, but HPE’s machine learning development system is directly targeting a current pain point, making it the most interesting announcement of the two.
“It’s almost an aaS [as a service] offer in your data center,” he said. “That’s what people are looking for by enabling AI model training in their business.