Data Science Investment Counter —Funding raised by UK data science companies in 2018.
£ 5.640 Million

Fujitsu develops deep learning acceleration technology

Fujitsu Laboratories have developed a new technology to substantially improve the speed of deep learning systems, the company revealed on Monday.

The newly achieved speed would be the worlds’ highest, according to the AI Bridging Cloud Infrastructure system, deployed by Fujitsu Limited for the National Institute of Advanced Industrial Science and Technology.

Deep learning applications have increased sharply in the last few years, from healthcare to robotics, from retail to the military. Because of this, the demand for efficient algorithms has accelerated by 30 times since 2017.

To calculate the speed of machine learning systems, tests are usually run against a deep neural network for image recognition called ResNet-50. The data is input into the system, and then training times are calculated using image data from the ImageNet Large Scale Visual Recognition Challenge 2012.

The new technology Fujitsu has developed should be able to expand computation volume per GPU without compromising training accuracy.

This would work through highly-efficient distributed parallel processing, achieved by the system by appropriately adjusting the learning rate following the degree of the deep learning training progress.

When this newly developed technology was applied to open source deep learning software using 2,048 GPUs in the ABCI system and measured for this benchmark, it beat the previous speed record by more than 30 seconds, completing the training in 74.7 seconds, the world’s highest speed.

Fujitsu Laboratories said they would try to increase the speed of deep learning further in the future, aiming to implement practical applications of this newly developed technology for Fujitsu’s servers and supercomputers.

The company was founded in 1968 as a subsidiary of Fujitsu. It is one of the leading research centres in the world. It has offices in Japan, China, the United States and Europe, and has a record of extensive research in areas such as Next-generation Services, Computer Servers, Networks, Electronic Devices and Advanced Materials.

Photo via Max Pixel.


Co-working space and blog dedicated to all things data science.

Subscribe to our newsletter