Fujitsu Doubles Deep Learning Neural Network Scale with Technology to Improve GPU Memory Efficiency
September 21, 2016 | FujitsuEstimated reading time: 4 minutes
Fujitsu Laboratories Ltd. today announced development of technology to streamline the internal memory of GPUs to support the growing neural network scale that works to heighten machine learning accuracy. This development has enabled neural network machine learning of a scale up to twice what was capable with previous technology.
Recent years have seen a focus on technologies that use GPUs for high-speed machine learning to support the huge volume of calculations necessary for deep learning processing. In order to make use of a GPU's high-speed calculation ability, the data to be used in a series of calculations needs to be stored in the GPU's internal memory. This, however, creates an issue where the scale of the neural network that could be built is limited by memory capacity.
Fujitsu Laboratories has now developed technology to improve memory efficiency, implementing and evaluating it in the Caffe open source deep learning framework software. Upon commencement of learning, the technology analyzes the structure of the neural network, and optimizes the order of calculations and allocation of data to memory, so that memory space can be efficiently reused. With AlexNet and VGGNet(1), image-recognition neural networks widely used in research, this technology was confirmed to enable the scale of learning of a neural network to be increased by up to roughly two times that of previous technology, thereby reducing the volume of internal GPU memory used by over 40%.
This technology makes it possible to expand the scale of a neural network that can be learned at high speed on one GPU, enabling the development of more accurate models. Fujitsu Laboratories aims to commercialize this technology as part of Fujitsu Limited's AI technology, Human Centric AI Zinrai, to work with customers in the use of AI.
Details of this technology were announced at MLSP (IEEE Machine Learning for Signal Processing 2016), an international conference held in Salerno, Italy from September 13 to 16.
Development Background
In recent years, deep learning has been gaining attention as a machine learning method that emulates the structure of the human brain. In deep learning, the more layers there are in a neural network, the more accurate it performs tasks, such as recognition or categorization. In order to increase accuracy, the scale of neural networks has been growing, but this lengthens learning times. Along with this, more attention is being placed on GPUs that execute computations with large volumes of data, and technology that accelerates the process by using multiple GPUs in parallel, as with supercomputers.
Issues
One method of increasing the scale of deep learning is to distribute a single neural network model across multiple computers and do the computations in parallel, but the volume of data that must be transmitted in exchanges between computers then becomes a bottleneck, greatly reducing learning speed. In order to take full advantage of the GPU's high-speed calculation capability, to the extent possible the data to be used in series of calculations needs to be stored in the GPU's internal memory. However, as GPU memory is usually smaller than that of an ordinary computer, there had been the issue of limitations in scale of neural networks capable of high-speed learning.
About the Newly Developed Technology
Now Fujitsu Laboratories has developed technology to streamline memory efficiency to expand the scale of a neural network for computations with one GPU, without using parallelization methods that greatly reduce learning speed. This technology reduces the volume of memory by enabling the reuse of memory resources; it takes advantage of the ability to independently execute both calculations to generate the intermediate error data from weighted data, and calculations to generate the weighted data error from intermediate data. When learning begins, the structure of every layer of the neural network is analyzed, and the order of calculations is changed so that memory space in which larger data has been allocated can be reused.
Effects
Fujitsu Laboratories implemented this newly developed technology into the Caffe open source deep learning software framework and measured the usage of GPU internal memory. In evaluations using AlexNet and VGGNet, which are widely used in research fields, it achieved reductions in memory usage volume of over 40% compared with before the application of this technology, enabling the scale of learning on a neural network for each GPU to be increased by up to roughly two times.
This will enable high-speed learning calculations using the full capability of a GPU, even with a large-scale neural network that requires complicated processing, accelerating the development of more accurate models.
Future Plans
Fujitsu Laboratories aims to commercialize this newly developed technology as part of Fujitsu Limited's AI technology, Human Centric AI Zinrai, by March 31, 2017. In addition, it plans to combine this technology with its already announced high-speed technology to process deep learning through GPU parallelization(2), and further improve these technologies.
(1) AlexNet and VGGNet: Multi-layered neural networks for image recognition.
In 2012 AlexNet received top honors in a competition for image classification, and in 2014 VGGNet received top honors in a competition for position detection, each achieving the highest recognition accuracy in the world. Today they each form the basis of image recognition neural networks.
(2) High-speed technology to process deep learning through GPU parallelization
"Fujitsu Develops High-Speed Technology to Process Deep Learning," .
About Fujitsu Laboratories
Founded in 1968 as a wholly owned subsidiary of Fujitsu Limited, Fujitsu Laboratories Ltd. is one of the premier research centers in the world. With a global network of laboratories in Japan, China, the United States and Europe, the organization conducts a wide range of basic and applied research in the areas of Next-generation Services, Computer Servers, Networks, Electronic Devices and Advanced Materials.
About Fujitsu Ltd
Fujitsu is the leading Japanese information and communication technology (ICT) company, offering a full range of technology products, solutions, and services. Approximately 159,000 Fujitsu people support customers in more than 100 countries. We use our experience and the power of ICT to shape the future of society with our customers. Fujitsu Limited (TSE:6702; ADR:FJTSY) reported consolidated revenues of 4.7 trillion yen (US$41 billion) for the fiscal year ended March 31, 2016.
Suggested Items
Warm Windows and Streamlined Skin Patches – IDTechEx Explores Flexible and Printed Electronics
04/26/2024 | IDTechExFlexible and printed electronics can be integrated into cars and homes to create modern aesthetics that are beneficial and easy to use. From luminous car controls to food labels that communicate the quality of food, the uses of this technology are endless and can upgrade many areas of everyday life.
iNEMI Packaging Tech Topic Series: Role of EDA in Advanced Semiconductor Packaging
04/26/2024 | iNEMIAdvanced semiconductor packaging with heterogenous integration has made on-package integration of multiple chips a crucial part of finding alternatives to transistor scaling. Historically, EDA tools for front-end and back-end design have evolved separately; however, design complexity and the increased number of die-to-die or die-to-substrate interconnections has led to the need for EDA tools that can support integration of overall design planning, implementation, and system analysis in a single cockpit.
Koh Young Showcases Award-winning Inspection Solutions at SMTconnect with SmartRep in Hall 4A.225
04/25/2024 | Koh Young TechnologyKoh Young Technology, the industry leader in True 3D measurement-based inspection solutions, will showcase an array of award-winning inspection and measurement solutions at SMTconnect alongside its sales partner, SmartRep, in booth 4A.225 at NürnbergMesse from June 11-13, 2023. The following offers a glimpse into what Koh Young will present at the tradeshow:
Real Time with… IPC APEX EXPO 2024: Plasmatreat: Innovative Surface Preparation Solutions
04/25/2024 | Real Time with...IPC APEX EXPOIn this interview, Editor Nolan Johnson speaks with Hardev Grewal, CEO and president of Plasmatreat, a developer of atmospheric plasma solutions. Plasmatreat uses clean compressed air and electricity to create plasma, offering environmentally friendly methods for surface preparation. Their technology measures plasma density for process optimization and can remove organic micro-contamination. Nolan and Hardev also discuss REDOX-Tool, a new technology for removing metal oxides.
Nanotechnology Market to Surpass $53.51 Billion by 2031
04/25/2024 | PRNewswireSkyQuest projects that the nanotechnology market will attain a value of USD 53.51 billion by 2031, with a CAGR of 36.4% over the forecast period (2024-2031).