Output Coding and Modularity for Multi-class Problems
نویسنده
چکیده
We study backpropagation networks learning classiication problems with multiple classes k > 3. The common way to code the output of a network is the one-per-class (OPC) method, where one bit is assigned to each class. A technique called error-correcting output coding (ECOC) converts the k-class learning problem into a large number of two-class learning problems. We propose to use modular architectures as a way to decorrelate the (redundant) network outputs. Various modular architectures are tested on an artiicial problem. We conclude that ECOC only improves upon OPC when combined with a suuciently modular approach.
منابع مشابه
Analyzing the Separability Matrix for ECOC coding
Error Correcting Output Codes (ECOC) have demonstrate to be a powerful tool for treating multi-class problems. Nevertheless, predefined ECOC designs may not benefit from Errorcorrecting principles for particular multi-class data. In this paper, we introduce the Separability matrix as a tool to study and enhance designs for ECOC coding.
متن کاملAdaptive Leader-Following and Leaderless Consensus of a Class of Nonlinear Systems Using Neural Networks
This paper deals with leader-following and leaderless consensus problems of high-order multi-input/multi-output (MIMO) multi-agent systems with unknown nonlinear dynamics in the presence of uncertain external disturbances. The agents may have different dynamics and communicate together under a directed graph. A distributed adaptive method is designed for both cases. The structures of the contro...
متن کاملMaximum Maintainability of Complex Systems via Modulation Based on DSM and Module Layout.Case Study:Laser Range Finder
The present paper aims to investigate the effects of modularity and the layout of subsystems and parts of a complex system on its maintainability. For this purpose, four objective functions have been considered simultaneously: I) maximizing the level of accordance between system design and optimum modularity design,II) maximizing the level of accessibility and the maintenance space required,III...
متن کاملData-driven decomposition for multi-class classification
This paper presents a new study on a method of designing a multi-class classifier: Data-driven Error Correcting Output Coding (DECOC). DECOC is based on the principle of Error Correcting Output Coding (ECOC), which uses a code matrix to decompose a multi-class problem into multiple binary problems. ECOC for multi-class classification hinges on the design of the code matrix. We propose to explor...
متن کاملImproving Multiclass Text Classification with Error-Correcting Output Coding and Sub-class Partitions
Error-Correcting Output Coding (ECOC) is a general framework for multiclass text classification with a set of binary classifiers. It can not only help a binary classifier solve multi-class classification problems, but also boost the performance of a multi-class classifier. When building each individual binary classifier in ECOC, multiple classes are randomly grouped into two disjoint groups: po...
متن کامل