Intellectual Property Portfolio

In this section, we introduce our intellectual property portfolio of Key Technologies (Computing, Network, AI, Data & Security, and Converging Technologies), which support the Key Focus Areas.

Computing

Fujitsu has been in the computer business since the 1950s. While many Japanese companies have introduced technology through partnerships with foreign companies, Fujitsu develops its own technology.
Fujitsu will continue to develop quantum and other computing technologies to realize innovation in the ICT infrastructure for supporting digital society. According to the Annual Report on Patent Administration 2022 (Statistics and Materials), Fujitsu ranked first in Japan in the number of patent registrations for I-6 computer technologies in the statistical table of registered numbers by field. The "Fugaku" supercomputer that has been in development by RIKEN and Fujitsu has successfully retained the top spot for five consecutive terms in multiple major high-performance computer rankings including HPCG and Graph500 BFS (Breadth-First Search).

• Japan Patent Office Annual Report 2022
• "Fugaku" supercomputer

Network

Fujitsu is committed to developing network software and cloud-native technologies based on the network technologies that we have cultivated over many years since the company's founding in 1935. Our goal is to make end-to-end virtualized cloud native networks available worldwide.
Fujitsu participates in and leads the activities of the O-RAN Alliance, an organization that promotes the expansion of next-generation wireless access networks, including 5G, by telecommunications carriers and equipment vendors, as well as the open deployment of such networks.
Fujitsu and NTT have embarked on a strategic alliance to realize the Sustainable Digital Society. Under this strategic alliance, the two companies will conduct joint research in fields that leverage their respective technological strengths, including optical technology, for which NTT and Fujitsu together boast of the world's highest number of patents, and other telecommunications technologies. Fujitsu is also working to strengthen its acquisition of standard-essential patents (SEPs) in order to promote uptake of our company's technologies in international markets. By participating in patent alliances and patent pools, we are also making efforts to collaborate with partners. By doing so, we aim to realize services that contribute to solving social issues by utilizing 5G and other technologies.

• Press release

AI

DeepTwin is an AI technology developed by Fujitsu for the first time in the world that accurately acquires essential features such as distribution and probability of high-dimensional data.
DeepTwin is a technology that applies Rate-Distortion theory to AI, which is the basis of technology for compressing video data to the minimum amount of information. In this study, we developed a new theory that by learning high-dimensional data with AI so that the amount of information can be most compressed, essential features such as the distribution and probability of the data can be acquired accurately. The outline is as follows.
DeepTwin uses an automatic encoder, which is one of the AI algorithms. An auto-encoder is a neural network trained to output the same data as the input data. The input data is dimensionally reduced before being output, and the restored data after dimensionality reduction becomes the output data. The performance of an auto-encoder is said to be better when the input data is reduced in dimension and the difference (error) between the input data and the output data is small. This is also similar to the concept of data compression described above.
In the case of dimension reduction in neural networks, especially by reducing the dimensions of high-dimensional data, the characteristics of the distribution of data in the data before the dimension reduction are lost, which causes performance degradation. Inspired by the Rate-Distortion theory used in data compression, Fujitsu was able to develop AI technology that can accurately capture the characteristics of data distribution in pre-dimensionality reduction data even after dimensionality reduction. A patent application was filed in 2019, and this technology was further refined and published in 2020. In 2020, DeepTwin was presented at the International Conference on Machine Learning 2020 (International Conference on Machine Learning 2020).
In reality, there are many more complex high-dimensional data and more difficult problems, but DeepTwin is continuing to develop technologies that can address more complex problems. As for DeepTwin, we will continue to apply for patents and present at academic conferences after the aforementioned publication and presentation, aiming to realize more practical AI.