An analysis of criteria for the evaluation of learning performance
暂无分享,去创建一个
The criteria for the evaluation of learning performance is essential for identifying a better learning algorithm. The basic criteria including accuracy and time complexity are commonly used in the evaluation of learning performance. The paper presents several new criteria including absolute LPA (low prediction accuracy) error, relative LPA error and predictive ability in addition to the various important criteria which are specific to the evaluation of learning performance in diverse learning task domains. The experimental results show that LPA error rates and predictive ability are useful in evaluating learning performance particularly in learning from large noisy databases.
[1] C. S. Wallace,et al. An Information Measure for Classification , 1968, Comput. J..
[2] Andrei N. Kolmogorov,et al. Logical basis for information theory and probability theory , 1968, IEEE Trans. Inf. Theory.
[3] Kevin B. Korb,et al. Causal Discovery via MML , 1996, ICML.
[4] C. S. Wallace,et al. Estimation and Inference by Compact Coding , 1987 .
[5] C. S. Wallace,et al. A General Selection Criterion for Inductive Inference , 1984, ECAI.