A novel confidence measure based on marginalization of jointly estimated error cause probabilities
نویسندگان
چکیده
We propose a novel confidence measure based on the marginalization of jointly estimated error cause probabilities. Conventional confidence measures directly score the reliability of recognition results. In contrast, our method first calculates joint confidence and error cause probabilities and then sums them with respect to the error cause patterns to obtain the marginal confidence probability. We show experimentally that, the confidence estimation accuracy obtained with the proposed method is significantly improved compared with that obtained with the conventional confidence measure.
منابع مشابه
A confidence-aware interval-based trust model
It is a common and useful task in a web of trust to evaluate the trust value between two nodes using intermediate nodes. This technique is widely used when the source node has no experience of direct interaction with the target node, or the direct trust is not reliable enough by itself. If trust is used to support decision-making, it is important to have not only an accurate estimate of trust, ...
متن کاملExact maximum coverage probabilities of confidence intervals with increasing bounds for Poisson distribution mean
A Poisson distribution is well used as a standard model for analyzing count data. So the Poisson distribution parameter estimation is widely applied in practice. Providing accurate confidence intervals for the discrete distribution parameters is very difficult. So far, many asymptotic confidence intervals for the mean of Poisson distribution is provided. It is known that the coverag...
متن کاملAn application of Measurement error evaluation using latent class analysis
Latent class analysis (LCA) is a method of evaluating non sampling errors, especially measurement error in categorical data. Biemer (2011) introduced four latent class modeling approaches: probability model parameterization, log linear model, modified path model, and graphical model using path diagrams. These models are interchangeable. Latent class probability models express l...
متن کاملEquivalent error bars for neural network classifiers trained by Bayesian inference
The topic of this paper is the problem of outlier detection for neural networks trained by Bayesian inference. I will show that marginalization is not a good method to get moderated probabilities for classes in outlying regions. The reason why marginalization fails to indicate outliers is analysed and an alternative measure, that is a more reliable indicator for outliers, is proposed. A simple ...
متن کاملEstimating parameters for probabilistic linkage of privacy-preserved datasets
BACKGROUND Probabilistic record linkage is a process used to bring together person-based records from within the same dataset (de-duplication) or from disparate datasets using pairwise comparisons and matching probabilities. The linkage strategy and associated match probabilities are often estimated through investigations into data quality and manual inspection. However, as privacy-preserved da...
متن کامل