A multi-scale non-parametric and parametric hybrid multi-category recognition algorithm with probabilistic outputs

作者：Lu Zhao^*; Fu Haoda; Song Gangbing

来源：Journal of Experimental & Theoretical Artificial Intelligence, 2015, 27(4): 487-500.

DOI：10.1080/0952813X.2014.977829

• kernel Fisher discriminant
• inter-class separability
• class-conditional density function
• multi-scale
• posterior probability
• binary tree

The synthesis of an effective multi-category nonlinear classifier with the capability to output calibrated posterior probabilities to enable post-processing is of great significance in practical recognition situations because the posterior probability reflects the assessment uncertainty. However, the estimation of posterior probability for multi-category classifiers is an unwieldy problem in the realm of pattern recognition, which usually is more intractable than that in dichotomic cases. In this paper, with the aid of binary tree representation for nested structures, a new polychotomous classification and posterior probability estimation scheme is developed on the strength of Bayesian decision theory. In particular, by capitalising on the intrinsic conexus between hierarchical structure and multi-scale analysis, the polychotomous multi-scale Bayesian kernel Fisher discriminant (KFD) is implemented for building the classifier at different scales for different levels. Finally, the performance of the proposed classification and posterior probability estimation algorithm is validated by designing a multi-category Bayesian KFD classifier for a benchmark satellite images dataset.

• 出版日期2015-7-4