Method for resource allocation among classifiers in classification systems

• US 7,401,062 B2
• Filed: 06/13/2006
• Issued: 07/15/2008
• Est. Priority Date: 06/13/2006
• Status: Active Grant

• Current Assignee: SINOEAST CONCEPT LIMITED (Tencent Holdings Limited)
• Original Assignee: International Business Machines Corporation
• Inventors: Lin, Ching-Yung, Smith, John R., Naphade, Milind R., Li, Chung-Sheng
• Primary Examiner(s): Hirl; Joseph P.
• Assistant Examiner(s): BUSS, BENJAMIN J
• Application Number:
  US11/451,972
• Publication Number:
  US 20070288408A1
• Time in Patent Office: 763 Days
• Field of Search: 706/14, 706/45, 706/48, 706/15, 706/25, 382/165, 709/223, 709/226
• US Class Current: 706/48
• CPC Class Codes: G06F 9/50 Allocation of resources, e.g. of the central processing unit [CPU]
• CPC Class Codes: G06N 20/00 Machine learning

Abstract

Methods and apparatus for optimizing resource allocation among data analysis functions in a classification system are provided. Each of the data analysis functions is characterized as a set of operating points in accordance with at least one of resource requirements and analysis quality. An operating point for each of the data analysis functions is selected in accordance with one or more constraints. The analysis functions are applied at selected operating points to optimize resource allocation among the data analysis functions in the classification system.

Claims

1. A method for optimizing resource allocation among data analysis functions in a classification system comprising the steps of:

characterizing each of the data analysis functions as a set of operating points in accordance with at least one of resource requirements and analysis quality;

selecting an operating point for each of the data analysis functions in accordance with one or more constraints; and

applying the data analysis functions at the selected operating points to optimize resource allocation among the data analysis functions in the classification system;

wherein the classification system comprises a plurality of classification stages, and each of the classification stages comprises one or more of the data analysis functions;

wherein the plurality of classification stages comprises at least one of feature extraction, atomic modeling and composite modeling;

wherein, in the step of characterizing each of the data analysis functions, the set of operating points correspond to classification algorithms represented in a multidimensional space with dimensions corresponding to at least resource requirements and analysis quality;

wherein, in the step of characterizing each of the data analysis functions, the resource requirements comprise at least one of computational, storage, communications and human requirements;

wherein, in the step of characterizing each of the data analysis functions, the analysis quality comprises at least one of average precision, classification rate, model validity assessments, receiver operating characteristic curves, reject curves, and precision curves;

wherein, in the step of characterizing each of the data analysis functions, the set of operating points results from an application of at least one of feature extraction, clustering, classification and visualization methods;

wherein the optimization occurs at one or more of the classification stages of the classification system; and

wherein the step of selecting an operating point comprises the step of defining a constrained optimization problem configured to perform at least one of (i) maximizing analysis quality with respect to at least one of available resources and response time; and

(ii) minimizing resource requirements with respect to a target analysis quality.