Saturday 19 July 2025
Delta debugging is a technique used by software developers and testers to identify the root cause of a program failure or error. It involves systematically removing code from a program until the problem disappears, revealing the faulty component. This process can be tedious and time-consuming, especially for complex programs.
Researchers at the University of New South Wales have developed a new approach called Probabilistic Monotonicity Assessment (PMA) that significantly accelerates delta debugging. PMA uses machine learning to dynamically model the program’s behavior during testing, allowing it to predict which code segments are likely to cause errors and eliminate them more efficiently.
The traditional method of delta debugging involves trying out different combinations of code removals until the problem is resolved. This approach can be slow because it requires a lot of trial and error. PMA, on the other hand, uses statistical analysis to identify patterns in the program’s behavior and make educated guesses about which code segments are causing errors.
The researchers tested their approach using two popular delta debugging tools: CHISEL and ProbDD. They found that PMA was able to reduce the processing time by 59.2% compared to CHISEL, accelerating the reduction process by 3.32 times, and decreasing the size of the final reduced programs by 6.7%. Against ProbDD, PMA reduced processing time by 22.0%, achieved a speedup of 1.34 times in the reduction process, and further decreased the sizes of the final reduced programs by 3.0%.
The potential benefits of PMA are significant. By reducing the time and effort required to identify the root cause of errors, developers can fix bugs more quickly and efficiently. This can lead to faster release cycles, improved software quality, and lower maintenance costs.
One of the challenges facing software development is the increasing complexity of modern programs. As codebases grow, it becomes more difficult to identify and isolate errors. PMA addresses this challenge by providing a more efficient way to debug complex programs.
The researchers plan to continue developing PMA, exploring ways to integrate it with other debugging techniques and improve its performance. They also hope to apply their approach to other areas of software engineering, such as testing and verification.
For developers, the implications of PMA are clear: faster debugging times, improved software quality, and reduced costs.
Cite this article: “Accelerating Delta Debugging with Probabilistic Monotonicity Assessment”, The Science Archive, 2025.
Debugging, Delta Debugging, Probabilistic Monotonicity Assessment, Machine Learning, Code Removal, Error Identification, Software Development, Bug Fixing, Complexity Reduction, Efficiency Improvement.
