基于随机奇异值分解的快速矩阵补全算法及其应用.pdfVIP

绗?*鍗 绗?鏈 璁＄畻鏈鸿緟鍔╄璁′笌鍥惧舰瀛﹀鎶 Vol. 2* No.* 201*骞?鏈 Journal of Computer-Aided Design Computer Graphics ***. 201* 鍩轰簬闅忔満濂囧紓鍊煎垎瑙ｇ殑蹇€熺煩闃佃ˉ鍏ㄧ畻娉曞強鍏跺簲鐢 1,2) 1,2) 1,2)* 3) 鍐 鏍 , 鏉庡彲娆 , 鍠绘枃鍋 , 榛庤€€鑸 1) (娓呭崕淇℃伅绉戝涓庢妧鏈浗瀹跺疄楠屽(绛? 鍖椾含 100084) 2) (娓呭崕澶у璁＄畻鏈虹瀛︿笌鎶€鏈郴 鍖椾含 100084) 3) (Department of Computer Science, Old Dominion University, Norfolk, VA 23529, USA) (yu-wj@) 鎽 瑕? 涓轰簡鍦ㄤ繚璇佺粨鏋滅簿搴︾殑鎯呭喌涓嬪姞蹇繍绠楅€熷害, 鏀硅繘浜嗙煩闃佃ˉ鍏ㄧ殑浠ｈ〃鎬х畻娉曗€斺€斿寮傚€奸棬闄?SVT)绠楁硶. 棣 鍏堝浜庤緭鍏ョ煩闃佃繘琛岃鏁村寲澶勭悊, 涔嬪悗鍦ㄦ瘡涓€姝ョ殑杩唬涓娇鐢ㄥ寮傚€煎垎瑙ｇ畻娉曞鐭╅樀杩涜鎭㈠. 鐢变簬姣忎釜杩唬姝ヤ腑 濂囧紓鍊煎垎瑙ｇ殑璁＄畻閲忓緢澶? 鏂囦腑鍊熼壌闅忔満鐭╅樀濂囧紓鍊煎垎瑙ｇ畻娉? 鎻愬嚭浣跨敤鍧楀厠闆锋礇澶凯浠ｈ繎浼煎寮傚€煎垎瑙ｇ畻娉曞拰瀛 绌洪棿澶嶇敤鎶€鏈殑蹇€烻VT 绠楁硶. 浣跨敤褰╄壊鍥惧儚鍜岀數褰辫瘎鍒嗙煩闃靛绠楁硶杩涜瀹為獙鐨勭粨鏋滆〃鏄? 蹇€烻VT 绠楁硶鍦ㄤ笉褰卞搷 鍥惧儚鎭㈠鍜岃瘎鍒嗘暟鎹娴嬫晥鏋滅殑鍚屾椂鏄捐憲鍦扮缉鐭簡璁＄畻鏃堕棿; 鍦ㄥ浘鍍忔仮澶嶅拰鐢靛奖璇勫垎棰勬祴鐨勫疄楠屼腑, 鍒嗗埆鍙栧緱浜嗛珮 杈?.1 鍊嶅拰3.2 鍊嶇殑鍔犻€熸瘮. 鍏抽敭璇岤 鐭╅樀琛ュ叏; 濂囧紓鍊煎垎瑙Ⅺ 濂囧紓鍊奸棬闄愮畻娉旣 闅忔満鐭╅樀绠楁硶; 鍥惧儚鎭㈠; 鎺ㄨ崘绯荤粺; 瀛愮┖闂村鐢 涓浘娉曞垎绫诲彿: TP391.41 Fast Matrix Completion Algorithm Based on Randomized Singular Value Decompo- sition and Its Applications Feng Xu1,2), Li Kexin1,2), Yu Wenjian1,2)*, and Li Yaohang3) 1) (TNList, Tsinghua University, Beijing, 100084) 2) (Department of Computer Science and Technology, Tsinghua University, Beijing, 100084) 3) (Department of Computer Science, Old Dominion University, Norfolk, VA 23529, USA) Abstract: The purpose of this work is to accelerate the singular value thresholding (SVT) algorithm, which is a representative algorithm of matrix completion, while preserving the accuracy. The algorithm first normalizes the input matrix, then use singular value decomposition(SVD) to complete the matrix. Because in each iteration step of the SVT algorithm the SVD of the data matrix is computed, the overall computational time of