The rewards of reusable machine learning code are becoming increasingly evident in the field of AI and data science. Nature Machine Intelligence has introduced a new article format called ‘reusability reports’ to showcase the value of high-quality code developments. These reports focus on testing the robustness, extendability, and reusability of previously published code, with a total of 12 reports already published.
Authors and referees have provided consistently positive feedback on the reusability reports, which undergo peer review and count as primary research articles. Unlike regular articles, reusability reports do not prioritize novelty but instead assess the technical correctness and value added with respect to the original work.
In a recent reusability report, Tao Xu et al. tested a bilinear attention model for predicting drug-target interactions and showcased its adaptability across various domains. Another example by Yingying Cao et al. demonstrated how code from the PENCIL method can be re-used to predict responses to immune checkpoint blockade therapy in skin cancer datasets.
Furthermore, Yuhe Zhang et al. revisited a deep learning method for reconstructing holographic images and extended it to non-perfect optical systems, incorporating system-specific response functions in the process. These examples illustrate the power of reusable code in advancing research and pushing the boundaries of machine learning applications.
