ENCORE introduces a novel G&V technique using ensemble learning on convolutional NMT models for automatic program repair. It outperforms traditional LSTM approaches by fixing 42 bugs across popular benchmarks, including bugs not fixed by existing techniques. The method is applicable to Java, C++, Python, and JavaScript, showcasing its versatility in fixing diverse bugs.
Automated program repair techniques rely on hard-coded rules but struggle with adapting to different programming languages. ENCORE's ensemble approach using convolutional NMT models overcomes this limitation by capturing diverse bug fixes and generating patches independently of context. The evaluation on popular benchmarks demonstrates the effectiveness of ENCORE in fixing a wide range of bugs across multiple programming languages.
The study highlights the importance of leveraging deep learning approaches like ensemble learning with convolutional NMT for automatic program repair. ENCORE's success in fixing complex bugs showcases its potential for improving software reliability and productivity in engineering tasks.
To Another Language
from source content
arxiv.org
Дополнительные вопросы