In Sparsely connected and Fully connected (Part 2), we described the characteristics and current status of the ASIC and MA versions, respectively. In this section, we will look back at the roles that hardware and software have played in the shift from the dawn of annealing machines to their widespread use, and also consider the roles they will play in the future. With a bit of personal insight, I will explain what hardware engineers expect from hardware.
Looking back at the world trend, D-Wave's quantum annealing machine was announced in 2011, followed by many annealing machines implemented in dedicated hardware, such as NTT/NII's coherent Ising machine and Hitachi's CMOS annealing machine, which have led the research and development in this field. These machines have been the driving force of research and development in this field. The common feature of these annealing machines is that they are hardware implementations of the Ising model, using physical devices such as quantum or optical oscillators or semiconductor CMOS as spins.
Subsequently, however, annealing machines that implement the fully connected Ising model have been announced which focus on the ease of use to various different problems; Hitachi's MA version of CMOS annealing is one such example. Fully connected annealing machines from various companies have been realized using software, and are now actually starting to be applied to social issues. With software, if you have an idea, you can simulate what you want to do using programming languages and algorithms, just like a person manipulating words and letters. Tuning makes it even easier to use and more flexible to respond to various social issues, bringing tangible value to people.
Machines implemented in hardware have various restrictions that make them difficult to use. In order to test what you want to do with hardware, you need a prototype of a device implemented in hardware, a system and equipment to run it, and changes to the facilities and environment for trial and error testing, so it takes an enormous amount of time and money from implementation to verification. Furthermore, the hardware itself has many limitations. For example, it is difficult to create a large number of couplings in an annealing machine, and increasing the number of bits in a coefficient also requires a lot of design effort.
However, even with such hard work, once hardware is completed, it can sometimes produce overwhelming performance that surprises people. This is because, compared to software, which executes complex instructions via CPUs and memory devices, hardware processing is faster and at the same time more energy-efficient because it is executed in a dedicated configuration.
Hardware engineers who have firsthand experience with these characteristics of hardware believe in its potential, so they have repeatedly innovated its performance to make it faster, finer, larger, or smaller, even when there are limitations. Moreover, it is precisely because of these limitations that there is a great sense of accomplishment when issues are overcome and tremendous performance is realized.
The Hitachi experiment of letting semiconductors perform the phenomenon of annealing also started from the infatuation of an engineer who strongly believed in the potential of hardware. Just as the creators of the first semiconductors did not anticipate the emergence of smartphones, new hardware has the chance to spawn unimaginable technological innovations
Annealing Cloud Web allows you to use CMOS annealing machines implemented with dedicated semiconductor ASICs, but even if you try to build an application using them, you may have difficulty creating problems or finding issues that fit, due to constraints on coupling and other factors. There may be some problems that cannot be created, or there is no issue that fits.
After having tried to run an application of signal control optimization for congestion relief, did you think of any other area of application? Or did you desire to use other implementations of annealing machines? Or were you interested to the idea of developing hardware that was more advanced than the ASIC version of the annealing machine? Both innovating with software and innovating with hardware are important missions in this field.
In the field of annealing machines, the first hook was the hardware realization of a quantum computer by D-Wave, which stimulated research and development. Later, a software approach was introduced, and it is expected that both approaches will evolve in an intertwined manner in the future. We can expect to see more and more examples of research and development that brings new value to the world through the coordination of these various layers.
With the emergence of domain-specific hardware, it is expected that people will create technology which can surpass it, and that the development of hardware and software will become more active as they influence each other. This will promote research and development in the field.
Combinatorial optimization will be explained by using two easy-to-understand examples of familiar combinatorial optimization problems. The importance of finding and using fast solution methods for combinatorial optimization processes in an IoT society will also be discussed.
In the wake of the COVID-19 epidemic, we present a use case in which an annealing machine is used to create shifts for researchers under constraints that take into account the risk of infection.
This article presents a use case in which an annealing machine is used to formulate a reinsurance portfolio that leverages the vast amount of data held by an insurance company to address natural catastrophe risks.
For beginners, this article explains the procedures and key points of the process of solving optimization problems with an annealing machine. The process involves dividing the problem into four stages: "organizing problem and defining problem," "formulation," "input data preparation," and "executing CMOS annealing machine".
This article provides an easy-to-understand explanation of the mathematical assumptions you need to understand, focusing on "quadratic" and "discrete," which are some of the characteristics of the problems that annealing machines excel at solving.