In recent years, there has been a lot of talk about Artificial Intelligence (AI), essentially Machine Learning (AA, or Machine Learning, ML). And yet AA has its roots (among others) in a discipline that is relatively unknown to the general public, but also very often to AI experts: operational research (OR).
A renewed interest in artificial intelligence
Since 2010, artificial intelligence (AI) has been the new enfant terrible. Not just any AI technology, but machine learning (ML) in particular.
The reason for this renewed interest is essentially due to the emergence of Deep Learning (DL, deep neural network or deep learning), a technique already developed in the 1960s. It has had its ups and downs, but at From the 1990s, the results began to be significant. In particular, the development and use of the backpropagation algorithm (see a detailed history of this algorithm here and here) allowed to train larger neural networks. And then of course, the advent of GPUs, etc.
Let’s move to the 2010s and 2020s and now it looks like the ML and DL in particular will solve everything (see Hinton’s nice claim here). And indeed, corporations and states are literally pouring billions of dollars into the development of AI, or more specifically, ML.
An AI train can hide another…
Most people have this image in mind when they talk about data science, big data and artificial intelligence: There is at least one important player missing here: operations research (OR). A much more realistic image would be:
Indeed, all of these areas are in some way based on OR. But what is operations research?
Operations research is an area of applied mathematics that can be described as (the?) science of optimization (and in fact is much more than that). If you’re interested (and if you consider yourself a serious data scientist, you really should be!), take a look at the page Wikipedia.
In fact, if you’re thinking about optimizing, you really should be thinking about OR. We will come back to this later, in another post. But it is not surprising that mathematical optimization plays an important role in all analytical approaches.
Now there’s…a third AI train coming! It does not ignore AA or OR, it combines them! It combines the strengths of these two areas and can compensate for the weaknesses of one with the strengths of the other. At Funartech, world pioneer in this combination, it is called “hybridization of ML and OR” . More and more people, companies and organizations are exploring this path. In Quebec, IVADO calls it Digital Intelligence (IN). We have come a long way. I remember 5 or 6 years ago when I started advocating for the combination of the two, very few people I met could grasp what it really meant. Even today there are many people who don’t understand or realize that combining not only makes sense, but is actually the strongest approach we have today (today , but not necessarily tomorrow! One idea to remember, however, is that of hybridization. Using several domains together, regardless of which ones, regardless of the problem, really allows us to go further!) to solve complex (industrial) problems. However, there are more and more convincing cases to demonstrate it, not to mention that this combination (in fact there are at least 4 types of ML/OR combinations) makes it possible to do things that are impossible in ML or OR only.
Operational research, a newcomer, really?
If OR is so important, how come so few people have heard of it? How can this be possible?
A few years ago, ML was considered a subfield of OR. Indeed, ML uses OR to optimize its predictions. 5-10 years ago, it was not even a debate and (virtually) everyone accepted this parentage. This is why, for example, Pr. Yoshua Bengio is a professor at … DIRO, the computer science and operational research department of the University of Montreal.
Personally, I always see things that way. I don’t want to reduce ML to using OR to optimize its predictions, because ML is so much more than that, but the main idea of developing algorithms that figure out for themselves what to do (that is- i.e. that don’t follow the if-then-else rules (even if ultimately everyone codes the algorithms with the same computer languages…) comes from OR and is in no way specific to ML or even at the OR.
OR is a well-established field that has seen great success since World War II. It is still unmatched today (not even by quantum computing, see for example Conquering the challenge of quantum optimization) when it comes to optimization. In another post, I’ll explain why ML is really bad for optimization in general. ML can be seen as a success stemming from OR and certainly an indication that optimization is a powerful engine for finding solutions to complex problems.
We would like to give thanks to the author of this post for this incredible material
Artificial intelligence: One train can hide another… – Actu IA
We have our social media profiles here as well as additional related pages here.https://www.ai-magazine.com/related-pages/