Skip to main content

IBM is cutting deep-learning processing times from days down to hours

deep learning
Daniel Kaesler/123RF
Deep learning uses algorithms inspired by the way human brains operate to put computers to work on tasks too big for organic gray matter. On Monday, IBM announced that a new record for the performance of a large neural network working with a large data set.

The company’s new deep-learning software brings together more than 256 graphics processing units across 64 IBM Power systems. The speed improvements brought about by the research come as a result of better communication between the array of GPUs.

Recommended Videos

Faster GPUs provide the necessary muscle to take on the kind of large scale problems today’s deep-learning systems are capable of tackling. However, the faster the components are, the more difficult it is to ensure that they are all working together as one cohesive unit.

As individual GPUs work on a particular problem, they share their learning with the other processors that make up the system. Conventional software is not capable of keeping up with the speed of current GPU technology, which means that time is wasted as they wait around for one another’s results.

Hillery Hunter, IBM’s director of systems acceleration and memory, compared the situation to the well-known parable of the blind men and the elephant. The company’s distributed deep-learning project has resulted in an API that developers can be used in conjunction with deep-learning frameworks to scale to multiple servers, making sure that their GPUs remain synchronized.

IBM recorded image recognition accuracy of 33.8 percent on a test run using 7.5 million images from the ImageNet-22K database. The previous best-published result was 29.8 percent, which was posted by Microsoft in October 2014 — in the past, accuracy has typically edged forward at a rate of about one percent in new implementations, so an improvement of four percent is considered to be a very good result.

Crucially, IBM’s system managed to achieve this in seven hours; the process that allowed Microsoft to set the previous record took 10 days to complete.

“Speed and scalability, which means higher accuracy, means that we can quickly retrain an AI model after there is a new cyber-security hack or a new fraud situation,” Hunter told Digital Trends. “Waiting for days or weeks to retrain the model is not practical — so being able to train accurately and within hours makes a big difference.”

These massive improvements in terms of speed, combined with advances in terms of accuracy make IBM’s distributed deep-learning software a major boon for anyone working with this technology. A technical preview of the API is available now as part of the company’s PowerAI enterprise deep-learning software.

Brad Jones
Former Digital Trends Contributor
Brad is an English-born writer currently splitting his time between Edinburgh and Pennsylvania. You can find him on Twitter…
The best inkjet printers of 2024: tested and reviewed
The EcoTank ET-3850 is a rare printer that lets me print envelopes from my phone.

Inkjet printers are incredibly versatile, printing text documents, labels, color graphics, and pictures in great detail. Some models are good enough to create professional photographic-quality prints. All-in-one inkjet printers add scan, copy, and fax, making them a good choice for small businesses and home offices.

We've reviewed inkjets from the best printer brands and collected our top recommendations to help you find a reliable solution that fits your needs for the best printer to buy.
The best overall inkjet printer: Canon MegaTank Pixma G3270

Read more
Apple defends the M4 Mac mini’s power button
The underside of the M4 Mac mini, showing its vent and power button.

Apple announced a new wave of product refreshes recently, and not only does the charging port for the Magic Mouse remain on the bottom of the device -- the M4 Mac mini's power button has been moved to the bottom, too. These design choices have riled up plenty of people, but it seems Apple stands by its new power button placement for the Mac mini.

In a video posted on Chinese social media platform Bilibili, Apple's Greg Joswiak not only defends the decision but praises it. He calls it a "kind of optimal spot for a power button," claiming that you just need to "kinda tuck your finger in there and hit the button."

Read more
Is AI already plateauing? New reporting suggests GPT-5 may be in trouble
A person sits in front of a laptop. On the laptop screen is the home page for OpenAI's ChatGPT artificial intelligence chatbot.

OpenAI's next-generation Orion model of ChatGPT, which is both rumored and denied to be arriving by the end of the year, may not be all it's been hyped to be once it arrives, according to a new report from The Information.

Citing anonymous OpenAI employees, the report claims the Orion model has shown a "far smaller" improvement over its GPT-4 predecessor than GPT-4 showed over GPT-3. Those sources also note that Orion "isn’t reliably better than its predecessor [GPT-4] in handling certain tasks," specifically coding applications, though the new model is notably stronger at general language capabilities, such as summarizing documents or generating emails.

Read more