4 結 語
為了解決通用行為模型在麵對新用戶時的不適用問題,本節提出了TrELM算法實現遷移學習。該方法是一種基於參數遷移的方法,是通過對ELM的目標函數進行修改,引入一個可以表示兩域差異的遷移學習量,實現ELM模型的遷移學習。利用TrELM算法實現通用模型的遷移,首先利用ELM分類器構建通用行為識別模型,可以得到源域中識別模型的輸出權值向量βS;之後通過對新用戶的少量行為樣本進行學習,修改通用模型的輸出向量為βt,實現對通用模型的修改,完成具有遷移學習功能的行為識別模型。實驗在真實數據集上進行,結果表明,該模型可以有效提高新用戶的行為識別正確率。
參考文獻
[1]張爍,段富.基於智能移動平台的情景感知技術研究[J].計算機應用與軟件,2013,30(8): 166-169.
[2]徐川龍,顧勤龍,姚明海.一種基於三維加速度傳感器的人體行為識別方法[J].計算機係統應用,2013,22(6):132-135.
[3] Dai Wenyuan,Yang Qiang,Xue Guirong,et al.Boosting for transfer learning[Z]. Proceedings of the 24th international conference on Machine learning ACM International Conference Proceeding Series, 2007:193-200.
[4] Dai W,Chen Y,Xue G,et al.Translated Learning:Transfer Learning across Different Feature Spaces[Z].In Advances in Neural Information Processing Systems,2008.
[5] Lawrence N D,Platt J C.Learning to Learn with the Informative Vector Machine[Z].In Proceedings of the International Conference in Machine Learning,2004.
[6] Bonilla E V, Chai K M A, Williams C K I. [Z].Nips,2008.
[7] Schwaighofer A,Tresp V,Yu K.Learning Gaussian Process Kernels via Hierarchical Bayes[J].In Advances in Neural Information Processing Systems, 2004: 1209-1216.
[8] Evgeniou T,Pontil M.Regularized multi-task learning[J].In Proceedings of 17th SIGKDD Conf[J]. on Knowledge Discovery and Data Mining,2004,28(1):109-117.
[9] Mihalkova L,Huynh T,Mooney R J. Mapping and revising Markov logic networks for transfer learning[C].National Conference on Artificial Intelligence-volume:AAAI Press,2007,608--614.
[10] Davis J, Domingos P. Deep Transfer via Second-Order Markov Logic[J]. Proceedings of the Aaai Workshop on Transfer Learning for Complex Tasks, 2009.