As digital transformation advances in society and in business, there are increasing needs for computers that can process massive volumes of sophisticated, complex data at high speeds. Amidst these developments comes a movement to harness quantum computing technology.
[Fujitsu Forum 2019 Business Session Report]
This session featured a conversation with Fujitsu’s Masaharu Yoshiyama and representatives from Fujitsu business partners, Japan Post’s Shigenori Nakaya and Toray’s Ryuji Tanimura. In addition, Waseda University Professor Nozomu Togawa, who is engaged in collaborative research, joined the stage alongside TC3’s Yoshihito Sudo, who works as a technologist, to talk about uses and potential applications of quantum computing technology.
Conventional Computers Hit the Limit for Quickly Solving Problems That Are Advanced, Complex, and Massive in Volume
VP, Platform Division, AI Services Business Unit
Digital Annealer first appeared in 2018 as an architectural solution for solving combinatorial optimization problems. Roughly one year later, Fujitsu has already received many inquiries from companies and organizations. Numerous case studies have emerged from these developments.
Why did Fujitsu develop Digital Annealer? To answer this question, we must start from the fact that conventional computers have reached their limit.
As digital transformation advances in society and in business, there are increasing needs for computers that can process massive volumes of sophisticated, complex data at high speeds. This has led many companies to invest in efforts to develop quantum computers. However, quantum computers are not yet developed enough for practical use, and it is predicted that it will take another 10 to 20 years before achieving sufficient practicality. Some often raised challenges include scalability and the fact that it is difficult to maintain quantum states.
Meanwhile, the amount of data in the world is increasing at an explosive speed. Fujitsu developed Digital Annealer as a computing solution to address these challenges.
Digital Annealer is the name of a new computer architecture that solves combinatorial optimization problems at high speed via a digital circuit that is based on the concept of quantum phenomena. There are two models of quantum computing technology: the quantum gate model and the Ising machine model. Digital Annealer is a type of machine specialized for combinatorial optimization problems using the Ising machine model. It achieves stable operation using a digital circuit, and it is easy to miniaturize because it employs semiconductors.
Digital Annealer, the new architectural solution for solving combinatorial optimization problems at high speed
Combinatorial optimization problems refer to the process of finding an optimized solution out of a set of given combinations. A well-known example is the traveling salesman problem. If a salesperson has five cities to visit, there are 120 possible orders for doing so. If there are 32 cities to visit, the number of combinations increases exponentially. Calculating such combinations takes ordinary computers a very long time. The process involves high electricity costs and is inefficient overall. Now, using Digital Annealer, such problems can be solved instantly.
It is sometimes said that only a very limited range of combinatorial optimization problems can be solved. However, it is known that solutions can potentially be applied to every single field. Particularly worthy of note is that solutions can be applied to both existing business fields and to new frontiers of business.
Existing fields include demand forecasting and chemical compound combination, which are fields that have heretofore been handled manually or by supercomputers. Regarding new frontiers of application, fields include advanced medicine, autonomous driving, and the development of advanced materials, which in the past could not be handled even with supercomputers.
However, the fact remains that new technologies struggle to permeate the market. Fujitsu's plan is to focus on collaborating with entities outside the company, including other companies and university research institutions, in order to encourage interest in Digital Annealer.
Advancing the Use of Digital Annealer by Collaborating with Companies and Universities
After Yoshiyama's speech, efforts led by Fujitsu business partners - including companies, universities, and organizations - were introduced. First, Japan Post’s Shigenori Nakaya introduced the company’s efforts to optimize truck dispatching for shipping deliveries.
Japan Post's success in reducing operation times by about 30% by optimizing vehicle dispatch
Exective Manager, Transport Division
Japan Post operates in three main segments: the postal and domestic logistics business, the post office business, and the international logistics business. Among these three businesses, the Transport Division handles transport tasks related to postal items in the postal and domestic logistics business.
Our customer's mail and packages are transported from the post office where they are received to the regional station that covers the relevant area. There are currently 62 regional sorting offices in Japan. Items are classified at the regional sorting offices, after which they are shipped to the regional station that covers the destination via airplane, truck, or rail. This is called mainline transport. Then, at the regional sorting office that covers the delivery destination, the items are classified for each collection and delivery post office and sent to collection and delivery post offices. Postal items are shipped along these three transport routes.
In mainline transport, if there are 4 bases, a total of 12 trucks is needed to connect all areas. If there are 62 bases, calculation requires 3,782 trips, but in practice, areas with low numbers of packages are serviced by relaying deliveries, so the fixed number of established trucks per day is approximately 3,000.
Including deliveries within the regions at each end of mainline transport, we operate a total of over 10,000 trucks per day. The total distance traveled on a daily basis is 1.72 million kilometers, which is equivalent to circling the earth 43 times.
The number of truck drivers in Japan peaked in 1995 at 980,000. By 2015, the number had dropped to 767,000. Meanwhile, the number of packages handled by Japan Post has increased 3.5-fold from 1.48 billion to 5.13 billion. We now face the problem of a shortage of truck drivers and an increased number of packages to deliver.
To solve this problem, we have been running the Post Logitech Innovation Program since 2017 as part of our efforts to innovate in Japan's distribution industry through co-creation with various companies that have innovative ideas and technologies.
In the second event, which we held in 2018 amidst the truck driver shortage, we focused on optimizing the shipping timetables of transport trips between post offices. Because this involves massive amounts of data, ordinary computers cannot perform the necessary calculations. As a solution, we selected a project developed by A*Quantum, a startup that specializes in software development technology for quantum computers. With help from Fujitsu, we worked with Fujitsu Laboratories to conduct demonstration experiments to optimize transport networks using Digital Annealer.
This was limited to a specific area in eastern Saitama Prefecture, where there are about 30 post offices, but during specific hours for a limited number of transport routes, based on data sets at each post office, we could derive the optimal routes in terms of transport loads, transport times and distances, and restrictions such as arrival times.
We conducted simulations on the optimized routes that we derived, and we discovered that we could reduce the number of trucks from 52 to 48 (about 8%), reduce costs by 7%, and improve carrying capacity by 12%. We also learned that we could reduce operation times by about 30% to reduce the amount of time drivers spend working, enabling us to reallocate their time to other tasks.
Going forward, we plan to lift the time of day restrictions and run the optimization simulation throughout the day to transport mail within the region. Ultimately, we want to tie these experiments to optimization of mainline transport, which takes much time and taxes drivers the most.
In drug discovery research, Toray reduced the estimated number of calculations for protein structures to 1/1,000th
In the next segment, Toray’s Ryuji Tanimura introduced the company’s efforts to use Digital Annealer in drug discovery and biotechnology research.
Pharmaceutical Research Laboratory/Digital Life Science Group
Toray is well known for its textile, plastic, and carbon fiber businesses, but the company is also developing a life science business. In this endeavor, we are leveraging know-how in advanced materials for research and development of pharmaceutical products and medical equipment.
At the Pharmaceutical Research Laboratories, we developed three new drugs: human interferon β (biologic drug), PGI2 derivative for oral administration (synthetic drug), and a remedy for intractable itchiness (synthetic drug).
We used Digital Annealer to predict the most stable structure for protein side chains.
Protein has various important functions in our bodies, including in the transport, synthesis, and breaking down of substances as well as information transmission. Many drugs are designed to bind to proteins in our bodies in order to control their functions.
Such proteins are substances made of amino acids that are linked together in specific arrangements. Depending on the arrangement, they have specific structural forms and perform certain functions. The structural arrangement information of proteins is especially useful in designing drugs that bind to proteins and control their functions, and much research focuses on deciphering these structural arrangements. However, it is difficult to experimentally determine the structural arrangements with measuring equipment such as X-ray or electron microscopes. In many cases, making a determination is impossible. Even if you can make a determination, the process is extremely costly and takes several months at a minimum.
To solve this, we are making efforts to predict structural arrangements through a method in which we calculate the lowest energy state required for protein to achieve a stable structure. Protein is composed of the main chain and side chains, and the amount of energy is determined by each chain’s structure. In this experiment, we limited the target to side chains, and we used Digital Annealer to predict the combination that has the least amount of energy (the most stable) out of all flexible side chains in relation to the corresponding main chain structure. Specifically, we used another method to exclude some side chain combinations that would definitely not become stable structures, and then we calculated the optimal side chain structure out of a number of combinations totaling about 10 to the 100th power.
As a result, a process that previously took more than four hours was completed in 20 seconds. In addition, we could obtain calculation results for proteins that we could not in the past.
To predict the structural arrangement of proteins, which is important in drug discovery and biotechnology research, we limited the target to side chains, but we could very quickly obtain predictions for the most stable structure. Moving forward, we hope to make further use of Digital Annealer to improve prediction accuracy, and we want to apply it to drug discovery research.
Digital Annealer in Use at Waseda University and in the Global Developer Community
The scope of co-creation with Fujitsu utilizing Digital Annealer continues to widen. In the next segment, Waseda University Professor Nozomu Togawa explained about the collaborative research underway with Fujitsu.
Collaborative research between Fujitsu and Waseda University in five areas, including energy and finance
Faculty of Science and Engineering
On September 19, 2018, Fujitsu and Waseda University entered into a comprehensive cooperation agreement on activities related to Digital Annealer. This collaborative research aims to solve real-world combinatorial optimization problems using Digital Annealer, with goals ranging from exploring use cases to accumulating human resources from a wide variety of fields.
As a place to carry out this collaborative research, we established the Fujitsu Co-Creation Research Laboratory at Waseda University on campus. Waseda University is making university-wide efforts to strongly promote industry-academic cooperation; among these, we established a group of research institutes called the Green Computing Systems Research Organization that focuses on specific fields. In one research institute, the Research Institute for Next-Gen Computing, we support the research underway at the Fujitsu Co-Creation Research Laboratory at Waseda University.
Moreover, we accepted submissions on a wide range of topics for research using Digital Annealer among all those researched at Waseda University. Starting in April 2019, we launched application projects for Digital Annealer after selecting a total of five topics: four submissions from the Faculty of Science and Engineering, namely "energy resource aggregation," "molecular evolution research," "allosteric channel analysis of protein molecules," and "application of log analysis in advertisement and digital marketing," and one submission related to the field of finance from the Faculty of Commerce. Moving forward, we hope to help advance these projects.
Digital Annealer use cases pushed by 1.4 million technologists worldwide
In the session’s closing segment, TC3’s Yoshihito Sudo took the stage and introduced usage case studies conducted with 1.4 million technologists around the world.
TC3 is the Japanese service provider of Topcoder, which is the world's largest community of technologists, including more than 1.4 million software engineers and designers. Topcoder, which originated in competitive programming, attracts highly skilled members from all over the globe.
In this instance, Topcoder held a demonstration contest for solving difficult problems using Digital Annealer. A total of 442 people from 63 countries and regions worldwide participated. Participants learned about Digital Annealer’s unique usage methods, formulated the problems, and coded to solve highly difficult problems using Digital Annealer.
Specifically, this challenge concerned complex algorithms called maximum cut problems. These crucial algorithms are used in a variety of fields, and many solutions have been published in theses. By entering actual data into the submitted programs, we could compare the results against the results described in past theses. Participants could use Digital Annealer to evaluate its capacity for developing practical algorithms and verify its effectiveness.
Going forward, we hope to continue to support those who are adept at problem-solving and who seek engineering resources and skills.
Image from the demonstration contest in which participants solve highly complex problems using Digital Annealer
Increasing the Speed to One Million Bits by 2020 and Accelerating Use in Actual Business
In closing, Fujitsu’s Masaharu Yoshiyama again took the stage to introduce case studies on the use of Digital Annealer in various fields.
For example, at NatWest Bank in Scotland, Digital Annealer is being used to calculate solutions to combinatorial optimization problems for asset portfolios at about 300 times the speed of conventional computers - and it does so much more accurately.
In addition, Fujitsu is engaged in various initiatives around the world, such as establishing a new company in Vancouver with the aim of accelerating the AI business, as well as expanding services in the US, Europe, and various regions in Asia.
Going forward, Fujitsu will continue to engage in development to further expand Digital Annealer’s scale, with the goal of reaching one million bits by 2020. Through expanded co-creation with companies and organizations, Fujitsu aims to further develop Digital Annealer and to provide more optimized services for actual business applications.
Executive Manager, Transport Division
Pharmaceutical Research Laboratory/Digital Life Science Group
Faculty of Science and Engineering
VP, Platform Division, AI Services Business Unit