PROJECT TITLE :

Efficient Parameter Estimation for Information Retrieval Using Black-Box Optimization - 2018

ABSTRACT:

Info Retrieval (IR) is that the complex of activities that represent data as information and rank the information representing data relevant to the user's data wants by a retrieval function. Such a perform involves parameters. They will in principle be set regardless of the precise set of documents and queries, but will in observe maximize retrieval effectiveness. But, algorithms to pick out retrieval operate parameters should be economical thanks to the big search space. We tend to can remark that: (i) all the tested strategies are similarly effective, but the plots of the maximum value of NDCG@20 at a given analysis show that our algorithm is additional economical; (ii) performance metrics and datasets studied in this Project seem to yield objective functions with few, if any, local optima with large basin of attraction; (iii) our algorithm is significantly a lot of economical, quickly finding parameterizations of the retrieval operate yielding high performance - abundant faster than line search.