Sign In

Communications of the ACM

Review articles

Model Learning

Model Learning, illustration

Credit: Marie Dommenget

We routinely manage to learn the behavior of a device or computer program by just pressing buttons and observing the resulting behavior. Especially children are very good at this and know exactly how to use a smartphone or microwave oven without ever consulting a manual. In such situations, we construct a mental model or state diagram of the device: through experiments we determine in which global states the device can be and which state transitions and outputs occur in response to which inputs. This article is about the design and application of algorithms that perform this task automatically.

Back to Top

Key Insights


There are numerous approaches where models of software components are inferred through analysis of the code, mining of system logs, or by performing tests. Many different types of models are inferred, for example, hidden Markov models, relations between variables, and class diagrams. In this article, we focus on one specific type of models, namely state diagrams, which are crucial for understanding the behavior of many software systems, such as (security and network) protocols and embedded control software. Model inference techniques can be either white box or black box, depending on whether they need access to the code. In this article, we discuss black box techniques. Advantages of these techniques are that they are relatively easy to use and can also be applied in situations where we do not have access to the code or to adequate white box tools. As a final restriction, we only consider techniques for active learning, that is, techniques that accomplish their task by actively doing experiments (tests) on the software. There is also an extensive body of work on passive learning, where models are constructed from (sets of) runs of the software. An advantage of active learning is that it provides models of the full behavior of a software component, and not just of the specific runs that have occurred during actual operation.


No entries found

Log in to Read the Full Article

Sign In

Sign in using your ACM Web Account username and password to access premium content if you are an ACM member, Communications subscriber or Digital Library subscriber.

Need Access?

Please select one of the options below for access to premium content and features.

Create a Web Account

If you are already an ACM member, Communications subscriber, or Digital Library subscriber, please set up a web account to access premium content on this site.

Join the ACM

Become a member to take full advantage of ACM's outstanding computing information resources, networking opportunities, and other benefits.

Subscribe to Communications of the ACM Magazine

Get full access to 50+ years of CACM content and receive the print version of the magazine monthly.

Purchase the Article

Non-members can purchase this article or a copy of the magazine in which it appears.