本論文以兩種演算法來檢驗動態調整 C 值在資料分類上的影響，探討其正確率和計算時間之表現。過去的研究顯示：(一)使用整體演算法能提高模型分類準確率。（二）將資料分批以分散式平行運算，能有效降低運算時間。（三）透過調整 C 值所得之分類器模型會有所不同。本論文共執行兩種實例分析演算法於數位化手寫數字資料庫(MNIST database)上。實例分析(一)演算法為 AdaBoostCascadeSVM.PL，結果發現動態調整 C 值，能有效地降低22~30程式運算時間，且 C 值為 25 與 50 時，能得到與無動態調整相近之分類準確率。實例分析(二)演算法為 AdaBoostCascadeRVM.PL，結果發現其演算法之核心 RVM 在處理大型資料因時間複雜度過高，有運算時間過長之問題。

In this thesis, we use two algorithms, AdaBoostCascadeSVM.PL and AdaBoostCascadeRVM.PL, which to verify the effect of classification with dynamic adjustment C value, and observe the performance of accuracy and computation time. In AdaBoostCascadeSVM.PL, classification with dynamic adjustment C value can save 22~30 computation time, and receive the similar accuracy when C value equal 25 and 50. On the other hand, the complexity of AdaBoostCascadeRVM.PL is too high to obtain classifier efficiently.

[1] B. Boser, I. Guyon, and V. Vapnik. A training algorithm for optimal margin clas- sifiers. In D. Haussler, editor, Proceedings of the 5th Annual ACM Workshop on Computational Learning Theory, pages 144–152. ACM Press, 1992.
[2] L. Breiman. Bagging predictors. Machine Learning, 24:123–140, 1996.
[3] L. Breiman. Random forests. Machine Learning, 45(1):5–32, 2001.
[4] R. Collobert, Y. Bengio, and S. Bengio. A parallel mixture of svms for very large scale problems. Neural Computation, 14(5):1105–1114, 2002.
[5] C. Cortes and V. Vapnik. Support-vector network. Machine Learning, 20:1–25, 1995.
[6] Y. Freund and R. E. Schapire. A decision-theoretic generalization of on-line learn- ing and an application to boosting. Journal of Computer and System Sciences, 55:119–139, 1995.
[7] H. P. Graf, E. Cosatto, L. Bottou, I. Durdanovic, and V. Vapnik. Parallel support vector machines: The cascade svm. Advances in Neural Information Processing Systems, 17:521–528, 2005.
[8] H. P. Graf, E. Cosatto, L. Bottou, I. Durdanovic, and V. Vapnik. Scalable and parallel boosting with mapreduce. IEEE Transactions on Knowledge and Data Engineering, 24(10):1904–1916, 2012.
[9] L. Hansen and P. Salamon. Neural network ensembles. IEEE Transactions on Pattern Analysis and Machine Intelligence, 12:993–1001, 1990.
[10] R. Maclin and D. Opitz. Popular ensemble methods: An empirical study. Journal Of Artificial Intelligence Research, 11:169–198, 1999.
[11] R. E. Schapire. The strength of weak learnability. Machine Learning, (2):197– 227, 1990.
[12] M. E. Tipping. Sparse bayesian learning and the relevance vector machine. Jour- nal of Machine Learning Research, 1:211–244, 2001.
[13] L. Xuchun, W. Lei, and S. Eric. Adaboost with svm-based component classifiers. Engineering Applications of Artificial Intelligence, 21:785–795, 2008.