This is the top of the page.
Displaying present location in the site.
  1. Home
  2. About NEC
  3. Research & Development
  4. Laboratories
  5. System Platform Research Laboratories
  6. Technology Fields
  7. Real-time analytics platform
Main content starts here.

Real-time analytics platform

Achieving the optimum operations required for social infrastructure in real time necessitates processing speeds that enable flexible control responsive to real world changes so that we can analyze the massive amounts of data (big data)—data exceeding the capacity of human intellect—obtained from sensors installed in countless locations.
We are therefore conducting R&D into high-speed analytics platform technologies that will enable us to smoothly execute extremely heavy processing flows with limited computer resources by acquiring large volumes of data from large numbers of sensors, using that data to analyze and predict the situation in the near future, and using those predictions to conduct optimal control of social infrastructures.

Distributed parallel processing platform

As data volumes skyrocket and demand for ever more advanced content analysis continues to rise, the length of time required for data analysis and machine learning is becoming an issue. NEC Central Research Laboratories is conducting research into a distributed parallel processing platform that applies high-performance computing (HPC) technology to solve this problem. This platform provides an interface resembling the similar middleware Spark, but that achieves far greater speeds by adopting communications speed and other innovations. In addition, it is also compatible with NEC's vector architecture, whose features can be utilized to achieve further speed improvements of data processing.

Profiling across spatio-temporal data (video retrieval technology)

Vast numbers of surveillance cameras are recently being used in urban and other areas. These cameras have been mainly used in criminal searches or other after-the-fact investigations—in other words, to find answers after an incident has occurred. Meanwhile, urgent problems such as terrorism and other forms of crime that can cause terrible damage to society, the field of crime prevention is beginning to attract a great deal of attention. NEC has developed a new video retrieval technology that applies face recognition technology to identify suspicious persons who are not preregistered in a database, such as people loitering on street corners, by establishing patterns of how and when they appear. Compared to the conventional method of simply searching for wanted suspects or pre-registered people, this technology is expected to be able to prevent crimes before they occur by enabling the early identification of potential criminals.

A demo screen of "Profiling across spatio-temporal data." The person who acted the role of a suspicious person was ranked as number one.
*These images are demo images used for development purposes.

Related Link

Technology utilizing vector processors

NEC has its own original vector processing technology and we are the only company in the world using it to develop processors, computing systems, and software. Conventional vector processors are used in large-scale numerical calculations such as weather simulations. We are conducting this R&D since we anticipate that vector processors will also be effective in social infrastructure and big data analysis applications. Our technology achieves speeds several times higher than those of conventional processors when executing image recognition, machine learning, and other advanced processing.

Technology utilizing FPGAs

As technologies continue to advance in the IoT era, we anticipate that data centers capable of conducting high-speed, advanced analysis will be required to meet the demand for executing huge numbers of computations under limited power resources. Against this background, attention is turning to the use of field programmable gate arrays (FPGAs) at data centers. FPGAs are a form of "soft" hardware that have a higher performance and are more power efficient than conventional CPUs. NEC is conducting research into platform technologies that will enable efficient communications between hardware and software, thereby making it possible to create a processor that closely couples an FPGA and CPU(s)—the kind of soft processor that is expected to feature prominently in the data centers ofnear future.

Related Link

Adaptive computing

NEC has developed and verified the world's first technology capable of controlling systems with dynamically configured hardware components (CPU, GPU, SSD, etc.) by using software, thereby enabling big data analysis to be performed with limited IT resources. In June 2014, Osaka University deployed a resource pool system that applies this technology, allowing them to fully utilize high-performance PCs remotely with no restrictions (see details). We have launched a consortium and are conducting other activities aimed to prevail this technology.

Brain-inspired computing

Various solutions for society utilizing AI technology are on the verge of fruition. NEC believes that, to realize those solutions, AI needs to have the following capabilities: the ability to correctly recognize situations not experienced in the past and make proper judgments; and the ability to rapidly make judgments according to changing situations. For AI to have those capabilities, edge computing infrastructure that processes information in close proximity to the scene is required, rather than the cloud concentrated computing infrastructure that is currently gaining mainstream acceptance. However, for this type of computing infrastructure to be realized, new architecture enabling AI to operate on much lower electrical power is imperative. NEC is aiming to realize next generation AI architecture that gives AI the capability of making judgments similar to the human brain by achieving "brain-inspired computing."
While realizing AI's sophisticated process of judging situations by reproducing human brain-like recognition, integration and analysis capabilities that imitate the human brain's macrostructure, we will reduce power consumption drastically by elaborately and directly reproducing the behavior of neurons and synapses in analog circuits that imitate the brain's microstructure. To tackle this ambitious feat, we will utilize NEC's accumulated computing system technologies, while collaborating with such leading research institutions in the field of brain-inspired computing as Osaka University and The University of Tokyo.

Related Links

Top of this page