Title :
Group-Based Active Query Selection for Rapid Diagnosis in Time-Critical Situations
Author :
Bellala, Gowtham ; Bhavnani, Suresh K. ; Scott, Clayton
Author_Institution :
Dept. of Electr. Eng. & Comput. Sci., Univ. of Michigan, Ann Arbor, MI, USA
Abstract :
In applications such as active learning and disease/fault diagnosis, one often encounters the problem of identifying an unknown object through a minimal number of queries. This problem has been referred to as query learning or object/entity identification. We consider three extensions of this fundamental problem that are motivated by practical considerations in real-world,time-critical identification tasks such as emergency response. First, we consider the problem where the objects are partitioned into groups, and the goal is to identify only the group to which the object belongs. Second, we address the situation where the queries are partitioned into groups, and an algorithm may suggest a group of queries to a human user, who then selects the actual query. Third, we consider the problem of object identification in the presence of persistent query noise, and relate it to group identification. To address these problems we show that a standard algorithm for object identification, known as generalized binary search, may be viewed as a generalization of Shannon-Fano coding. We then extend this result to the group-based settings, leading to new algorithms, whose performance is demonstrated through a logarithmic approximation bound, and through experiments on simulated data and a database used for toxic chemical identification.
Keywords :
approximation theory; identification; learning (artificial intelligence); query processing; Shannon-Fano coding; generalized binary search; group-based active query selection; logarithmic approximation bound; persistent query noise; query learning; rapid diagnosis technique; time-critical situation; toxic chemical identification; unknown object identification; Channel coding; Chemicals; Decision trees; Noise; Object recognition; Toxic chemicals; Active learning; Shannon-Fano coding; decision trees; generalized binary search; persistent noise; submodularity;
Journal_Title :
Information Theory, IEEE Transactions on
DOI :
10.1109/TIT.2011.2169296