摘要: We are concerned with the issue of detecting model changes in probability distributions. specifically consider strategies based on minimum description length (MDL) principle. theoretically analyze their basic performance from two aspects: data compression and hypothesis testing. From view compression, we derive a new bound minimax regret for changes. Here, mini–max is defined as worst-case code-length relative to least normalized maximum likelihood over all testing, reduce change detection into simple testing problem. thereby upper bounds error probabilities MDL-based test. The valid finite sample size related information-theoretic complexity well discrepancy measure hypotheses be tested.
sci-hub.se 参考文章(34)
O Bousquet, A Note on Parameter Tuning for On-Line Shifting Algorithms Max Planck Institute for Biological Cybernetics. ,(2003)
Jorma Rissanen, Optimal Estimation of Parameters Cambridge University Press. ,(2012) , 10.1017/CBO9780511791635
Eiji Takimoto, Manfred K. Warmuth, The Last-Step Minimax Algorithm Lecture Notes in Computer Science. pp. 279- 290 ,(2000) , 10.1007/3-540-40992-0_21
Indrė Žliobaitė, Learning under Concept Drift: an Overview arXiv: Artificial Intelligence. ,(2010)
Jon Kleinberg, Bursty and Hierarchical Structure in Streams Data Mining and Knowledge Discovery. ,vol. 7, pp. 373- 397 ,(2003) , 10.1023/A:1024940629314
João Gama, Pedro Medas, Gladys Castillo, Pedro Rodrigues, Learning with Drift Detection Advances in Artificial Intelligence – SBIA 2004. pp. 286- 295 ,(2004) , 10.1007/978-3-540-28645-5_29
P.A.J. Volf, F.M.J. Willems, Switching between two universal source coding algorithms data compression conference. pp. 491- 500 ,(1998) , 10.1109/DCC.1998.672217
Kenji Yamanishi, Yuko Maruyama, Dynamic syslog mining for network failure monitoring knowledge discovery and data mining. pp. 499- 508 ,(2005) , 10.1145/1081870.1081927
Mark Herbster, Manfred K. Warmuth, Tracking the Best Expert Machine Learning. ,vol. 32, pp. 151- 178 ,(1998) , 10.1023/A:1007424614876
Ciprian Doru Giurcăneanu, Seyed Alireza Razavi, None, AR order selection in the case when the model parameters are estimated by forgetting factor least-squares algorithms Signal Processing. ,vol. 90, pp. 451- 466 ,(2010) , 10.1016/J.SIGPRO.2009.07.011