loading page

Abstract Drafting
  • Faical Yannick Palingwende Congo,
  • Daniel Wheeler
Faical Yannick Palingwende Congo

Corresponding Author:[email protected]

Author Profile
Daniel Wheeler
Author Profile

Abstract

This research will investigate the concern of software execution reproducibility. Reproducibility in that sense denotes three aspects that can be assessed. Firstly dicussing the very reproducibile aspect of a software is certainly one of them. Secondly laying out a standard structure and terminology that will be the base to represent and reproduce an execution context. Finally macthing between an original context and a reproduced one to diccuss certain similarity metrics regarding the ouputs, meaning also how far away is the reproduced ouput from the original one then associate that drift to the context environmental change issues and define how to use it a knowledge.

In this research we tend more to lean toward the second and then the third aspects of this area. Currently there are some interests on that field because it becomes obvious for one doing more or less any research (simulation, experimentations, etc...) computer related that having a way to preserve the state of a trial for various reasons (sharing, publication, knowledge by difference or similarity, ...) will be more convenient that using current old fashion solutions (Lab note books, Text editors, etc...).

Existing tools like sumatra aret trying to provide a solution to that requirement. One may ask what is the conciliation part between those citied aspects. Well, one thing is to be able to have some records of software execution contexts and another thing is to be able to feature a way to compute the execution to that same very context and correlate the results with the original context.

Our final goal is to formalise the way of automating these two main aspects we are interested in and have this model implemented along with what already exists in Sumatra as a main case study and also some possible secondary case studies. The first apsect will probably be discussed and have some of our time dedicated to it at some point.

One remote reason as a use to this research is for AI, specially predictive behavior and globally pattern recognition. In fact, the history context of a program during its different life cycle can be used as a reflexive input to the same program or to another program. This will enable environmental changes on the host to be detected and its impact on the programs hosted. I will let you imagine the leverage of this kind of knowledge.