Dear all,
you are cordially invited to join the master thesis defense given by Syed Ahmed on the 27th of September on 17:00. You can find title, abstract and Zoom link below.
Best regards
Johannes
Title: Techniques for Candidate Selection in the Context of Interactive Machine Learning to Identify Irrelevant Files
Abstract:
The more a computer system is used, the more digital files and artifacts are produced. Sometimes those files are necessary and should be kept on the system - for example - the latest set of requirements from the customer in form of an excel file, or a PDF invoice for tax reasons. However, there is a large number of files, which are not needed beyond a certain period, and should be deleted. Normally, these files keep piling up, bloating our computers and data centers in the process. Often it is very difficult to motivate users in an organization to go through their files one by one and deleting the ones they do not need anymore. At the same time, the IT department cannot just clean up the whole drives in one sweep stroke, because that action could potentially have disastrous consequences. Although there are some solutions for cleaning machine-created clutter like bloated registry, browser cache, and cookies etc. for various operating systems, there are no commercially available solutions to help us sort out the mess that is created in the user-controlled workspaces outside of these designated temporary folders. In academia however, there have been efforts in this regard under the flag of a project called Dare2Del. It is a companion system based upon the principles of explainable AI. In this work, we expand upon the work done in this project by implementing a proof of concept prototype in a real-world company, and lay down the foundation for the employment of interactive machine learning with real data at the same time. Moreover, we also propose some imperative methods to select candidate digital objects for deletion in a non-greedy fashion.
Link: https://uni-bamberg.zoom.us/j/97277487878