Rhys Davies, Alec Jackson, Jason Boyd
Abstract
How should employers in the software field narrow a list of new graduate candidates from a computer science department at a university to include only the best-suited candidates that meet their criteria? We propose implementing a centralized agent that uses combinatorial pure exploration (CPE) to perform costly exploration of a set of arms of new graduate candidates. Employers will specify minimum criteria to the centralized agent, which will in turn explore the set of arms with a given number of candidates. We will examine what number of candidates must be given to each individual agent (or arm) to produce a candidate which meets the criteria set by the employer. We will also examine what portion of agents find a candidate from the computer science department at Utah State University given that threshold, and conclude what portion of Utah State University computer science students are qualified for real-world software positions.
Rough Program Steps
- The first pull will acquire the top-k utilities from the candidates.
- each piece of the candidate will be assigned a utility that coincides with their desired field of work.
- an overall utility will be calculated from those utility pieces that will be used for candidate selection.
- Each subsequent pull will randomly add or subtract a certain amount of utility (simulating employers learning about candidates).
- the second pull could possibly look at the employer's utility pieces and pull candidates who score higher with that employer.
- The last pull will be random as well (how well the candidate blends with the team).
Step 1: Generating Data
- Candidate Generation
- Employer Generation
- Utility Generation