Reproducibility challenge of the paper "code2seq: Generating Sequences from Structured Representations of Code" http://code2seq.org
Inspired by the reproducibility challenge encouraged by the International Conference on Representation Learning, this study attempts to contribute to the ongoing effort of addressing the lack of reproducibility of published results in Machine Learning papers. This paper aims to review the publication ‘Code2Seq: Generating Sequences from Structured Representations of Code’ by reimplementing it in a different framework, training it using the same datasets and performing a qualitative comparison of the model. The reproduced results successfully demonstrates an improvement over previous state-of-the-art results, but fail to achieve the quoted performance in the original publication.