Can you trust your experimental results?
Many contributions in computer science rely on quantitative experiments to validate their efficacy. Well-designed experiments provide useful insights while poorly-designed experiments can mislead. Unfortunately, experiments are difficult to design and even seasoned experimenters can make mistakes. This paper presents a framework that enables us to talk and reason about experimental evaluation. As such, we hope it will help our community to avoid and recognize mistakes in our experiments. This paper is the outcome of the Evaluate 2011 workshop whose goal was to improve experimental methodology in computer science.
- Login to post comments
Publication Highlights
OOPSLA'15 - Use at Own Risk
PPPJ'13 - Jikes RVM Debugger
PLDI'12 - Algorithmic Profiling
OOPSLA'11 - Catch Me
ECOOP'11 - Beauty and Beast
PLDI'10 - Profiler (In)Accuracy
ASPLOS'09 - Measurement Bias
More...
Blast
Our framework for bytecode-level information-flow tracing of Java programs.
Jikes RDB
Working with the Jikes RVM? Use Jikes RDB for debugging your VM hacks.
Now built on top of LLDB, so it works on OS X and on Linux.