Framework for a Citizen Science Research Using Machine Learning Methods

Modern science is characterized by the growing number of citizen research. These research, in turn, can be conducted by a large number of people, so citizen science demands a higher level of collaboration. Another key point is a reproducibility, which means one can accurately repeat the results of another scientist. The concept of reproducibility is of particular importance, when it comes to a rapidly developing field of knowledge. We consider machine learning as an example of such area. Machine learning scientists (or more generally data-scientists) use a massive set of tools (such as utilities, program libraries, etc.), and one researcher has to use the same set as another one uses to get the same, or at least compatible results. We present project Everware that combines git repository management system (GitHub, GitLab), virtualization technique (Docker) and a platform for conducting research in a collaborative manner (Jupyter) in order to bring together collaboration and reproducibility.