Story image

How Sony is using distributed learning to create enterprise AI

19 Nov 18

Sony announced that by utilising its deep learning development framework "Core Library: Neural Network Libraries" in addition to the AI Bridging Cloud Infrastructure (ABCI), a world-class computing infrastructure for AI processing, it has achieved the fastest deep learning speeds in the world.

Deep learning is a method of machine learning which uses neural networks modelled after the human brain. By harnessing deep learning, image and sound recognition capabilities have seen rapid growth in recent years, even outperforming humans in certain domains. 

However, the size of data used in this learning and model parameters used to improve recognition accuracy has been increasing, causing a subsequent rise in calculation times. 

In some cases, it has taken weeks or even months to conduct a single learning session. Because AI development requires a continuous process of trial-and-error, shortening this learning time is of the utmost importance. 

To this end, distributed learning using multiple GPUs as a means of shortening learning times is emerging as a popular solution.

When increasing the number of GPUs for distributed learning, there are cases where an increase to batch sizes (the amount of data to be processed at one time) halts the learning process and other cases where the learning speed actually decreases due to the processing delays caused by data transmission times between GPUs. 

By utilising technology that can determine the optimal batch sizes and the appropriate number of GPUs based on the current state of the learning process, Sony makes it possible to carry out learning even in large-scale GPU environments such as ABCI, and increased transmission speeds between GPUs through data synchronisation technology optimised for ABCI's system structure. 

These technologies were implemented into the "Neural Network Libraries," and used ABCI computing resources provided by AIST's "ABCI Grand Challenge" to carry out learning. 

As a result, it was able to complete ImageNet/ResNet-50*2 (the general industry benchmark used to measure distributed learning speeds for deep learning) in approximately 3.7 minutes (when using as many as 2,176 GPUs), achieving the world's fastest speeds to date. 

The results of this experiment demonstrate that learning/execution carried out using Neural Network Libraries can achieve world-class speeds and that by utilising the same framework, it is possible to conduct technology development using deep learning with a shorter trial-and-error period. 

Moving forward, Sony will continue development on related technologies and seek to contribute to the development of society using AI technology.

How Adobe aims to drive digital transformation for financial services
Digital transformation is a requirement for ongoing competitiveness that clearly helps businesses run more efficiently.
Using blockchain to ensure regulatory compliance
“Data privacy regulations such as the GDPR require you to put better safeguards in place to protect customer data, and to prove you’ve done it."
Human value must be put back in marketing - report
“Digital is now so widely adopted that its novelty has worn off. In their attempt to declutter, people are being more selective about which products and services they incorporate into their daily lives."
A10 aims to secure Kubernetes container environments
The solution aims to provide teams deploying microservices applications with an automated way to integrate enterprise-grade security with comprehensive application visibility and analytics.
DigiCert conquers Google's distrust of Symantec certs
“This could have been an extremely disruptive event to online commerce," comments DigiCert CEO John Merrill. 
Microsoft NZ bids Goldie a “fond farewell”
Microsoft New Zealand director of commercial and partner business takes new role across the Tasman. The search for his replacement has begun.
Google says ‘circular economy’ needed for data centres
Google's Sustainability Officer believes major changes are critical in data centres to emulate the cyclical life of nature.
One Identity a Visionary in Magic Quad for PAM
One Identity was recognised in the Gartner Magic Quadrant for Privileged Access Management for completeness of vision and ability to execute.