■ 일 시 : 2018. 2. 7(수), 5:00 PM ~ 6:00 PM
■ 장 소 : 산업경영동(E2-2) 멀티미디어실(2501)
■ 연 사 : 인하대 이우기 교수
■ 제 목 : Linear Programming Approaches for Graph Environments
One of the fundamental mechanisms for searching complex graphs is to grasp what the user’s intention and thereof by graph objects to support it effectively. The basic unit of search and retrieval in general keyword search over graph is a tree that connects all the query keywords hierarchically. The trees like a Steiner tree that include a set of specific keywords can be calculated by the weight from which the queries and the solutions are extracted. This new answering paradigm, in contrast to single results given by search engines, brings forth new challenges for ranking. A novel aggregation operator is deployed; to combine the content relevance based sets and, compute query dependent edge weights. The rank of an answer is computed by non-monotonic addition of edge weights, as per their relevance to keyword query. We define a unit of graphed nodes derived from the tree in order to focus on the user's interest in terms of keyword that the query requests. Even if a graph node does not include all the keyword, the related nodes connected by the links are derived and displayed onto various devices including desk top or the mobile browser. We adopted Linear Programming approaches so that various advantages developed in the industrial engineering communities so far can be nourished such as optimized structuring, sensitivity analysis, and dual models. The integration of the graph nodes to display Top-k results synchronously, which will be convoyed by an optimal approach comparing with Linear Programming models as well as typical meta-heuristics.
※세미나 마일리지 적용※