4/1/2019

Fujitsu Touts Faster Deep Learning

Facebook Share Icon LinkedIn Share Icon Twitter Share Icon Share by EMail icon Print Icon

Japan’s Fujitsu Laboratories Ltd. claims its latest artificial intelligence software enables the world’s fastest machine learning processing.

Share

Facebook Share Icon LinkedIn Share Icon Twitter Share Icon Share by EMail icon Print Icon

Japan’s Fujitsu Laboratories Ltd. claims its latest artificial intelligence software enables the world’s fastest machine learning processing.

Used in a 2,048 GPU network, the so-called deep learning system completed training calculations in less than 75 seconds—about 30% faster than the previous best. The gain was achieved in part by expanding computational volume per graphic processing unit used for image recognition, according to Fujitsu.

Deep learning is considered a key enabler for next-generation autonomous vehicles. The technology allows vehicles to quickly identify images/patterns and build a reference database that trains software to make decisions based on learned inputs and outcomes.

Fujitsu tested its system at Japan’s National Institute of Advanced Industrial Science and Technology. The technology was tested on Microsoft Corp.’s ResNet-50 deep-learning neural network.