Researching empirical fields in the current era involves working with data, code and software environments. Researchers are expected to guarantee high reproducibility of their results, which involves a complete description of all three ingredients. Due to the exceptional advances in infrastructural (hardware) and corresponding methodological (software, computer science, algorithms, methodologies) areas, the amount of feasible and applicable scientific methodologies and algorithms has increased substantially. This increase was partially caused and initiated by the Open Source Software (OSS) movement, which has now become a dominant means of scientific software development and distribution.
This course aims to demonstrate the application of OSS to tackle two complex research problems in Machine Learning. The main goals of the course are:
To demonstrate how an OSS environment can be set up and managed for a research project,
To demonstrate how Resolos and the OSS tool can manage the project lifecycle,
Provide examples of how tool choice informs methodology and vice versa,
and finally, to show a curated set of tools a researcher can rely on during critical steps of conducting a research project.
The goals are achieved by replicating the results of two selected, highly technical and computationally intensive research papers:
Chen, H., Didisheim, A. and Scheidegger, S., 2021. Deep Structural Estimation: With an Application to Option Pricing. arXiv preprint arXiv:2102.09209.
You, J., Ying, R., Ren, X., Hamilton, W. and Leskovec, J., 2018, July. Graphrnn: Generating realistic graphs with deep auto-regressive models. In International conference on machine learning (pp. 5708-5717). PMLR.