Machine learning systems are systems where associated code is heavily dependent on the accuracy of the implementation of complex mathematical equations. As a result, algorithm and method bugs are the most prevalent type of bug found in machine learning systems and these bugs are the most challenging bugs to fix in these systems. Automated verification of method implementation correctness would alleviate developers of time-intensive debugging for these bugs. As a first step, in this work, we propose a method that can be further developed to map mathematical equations from academic papers to their corresponding code. We also conduct a review of the nature and complexity of methods found in academic papers of machine learning systems. Results from our work demonstrate that representation generation algorithms like subword tokenization and models that directly enforce representational similarity show promise for future work in code matching. All code for this work can be found here.