Author: Edward Rothberg, PhD
There are, generally speaking, two types of people in the mathematical optimization software field:
- Optimization solver developers: The technical experts who devise and implement the algorithms that are the motor that makes every mathematical optimization application run.
- Optimization solver users: The operations research (OR) specialists, data scientists and businesspeople who design, deploy and use the mathematical optimization applications to help organizations tackle their complex, real-world problems and make optimal decisions.
Interestingly, these two groups operate separately for the most part, with the users often in the dark about how exactly the mysterious algorithmic alchemy inside the solver works and the developers largely unaware of the myriad mathematical optimization applications that are being implemented in the business world today.
As a mathematical optimization software veteran who has devoted my career to developing (in collaboration with my colleagues) the world’s fastest and most powerful mathematical optimization solver, I’m one of the developers.
Of course, I know all about the many high-profile use cases of mathematical optimization that have delivered billions of dollars in cost savings and revenue growth for leading companies across numerous industries, including financial services, telecommunications, electric power, manufacturing and logistics.
But almost every day I learn about a new, cutting-edge application that I’ve never heard of or even imagined was possible. I must admit that many times, after speaking with customers about their applications, I slap my head and say, “I didn’t know mathematical optimization could be used for that!”
There are so many innovative use cases of this AI technology, encompassing everything from food delivery dispatch to urban parking optimization to earthquake modeling to subscription box service optimization to scaffolding design optimization to energy asset modeling and many, many more. The possibilities are essentially endless.
But, you may ask, how can one single technology be used to address such a vast variety of business problems?
Here are three main reasons why I think mathematical optimization is such a powerful and versatile AI problem-solving technology:
1. The Broad Applicability Of The Mathematical Optimization Solver
For most users, a mathematical optimization solver functions as a black box: Although they can use the outputs of the technology, they have little understanding of its inner workings — but they really don’t need to. All they need to do is integrate the solver into their applications and then let it automatically crunch their data and crank out optimal solutions to their business problems.
Think of a solver (to elaborate on the analogy I made at beginning of this article) as being like an electric motor that can be inserted into so many different things (household appliances, ships, power tools, cars, etc.) and used to make them run. Over the years, as the number of uses of the solver (and electric motor) has increased, the performance of the technology has improved — and this phenomenon continues today.
Indeed, inside the black box is a collection of the world’s most sophisticated, state-of-the-art mathematical algorithms — the result of over 70 years of hard work and brilliant ideas by some of the leading minds in the industry.
The broad applicability of the black-box solver helps make mathematical optimization an AI technology with enduring and far-reaching impact in our world today.
2. The Flexibility Of Mathematical Optimization Models
A mathematical optimization application has two main components: the solver described in the section above, and a mathematical model that encompasses your unique, real-world business problem.
Although the solver is a standard, off-the-shelf software technology, the model is a highly customized and detailed mathematical representation of your business problem (which includes your company’s specific business rules, business goals and decision-making processes).
The flexibility of the model gives you the ability to capture your complex, mission-critical business problems in mathematical terms and, using a mathematical optimization solver, generate an optimal solution to those business problems.
This robust modeling capability is a key reason mathematical optimization remains such a versatile and pivotal technology for businesses today. No matter how complex or challenging your business problem is, if it can be defined as a mathematical model, it can almost certainly be solved using mathematical optimization.
3. The Capability To Make Optimal Decisions Based On The Latest Available Data
Data is the lifeblood of every AI technology. Some AI tools, such as machine learning, rely on historical data to make predictions, while other tools, such as mathematical optimization, leverage the latest available data and models of your present-day business environment to facilitate (and often automate) decision making.
You can reveal the value of your data by using mathematical optimization to automatically generate solutions to the real-world business problems you are facing today and make optimal business decisions.
Looking To The Future
The era of AI is upon us. Gartner predicts that by 2024, 75% of enterprises will have operationalized AI. The question is: Which technologies will remain relevant?
For the reasons highlighted in the article, I believe that mathematical optimization will continue to be an essential tool used in an ever-expanding array of applications by organizations to address their most challenging and critical business problems and make optimal decisions.
Indeed, the opportunities for mathematical optimization across various industries are virtually limitless. In the coming years, users in the business world will no doubt continue to build and deploy new, innovative applications, and I (along with others in the software development community) will continue to be amazed by these groundbreaking use cases.
This article was originally published on Forbes.com here.