计算机科学与探索 ›› 2022, Vol. 16 ›› Issue (8): 1809-1818.DOI: 10.3778/j.issn.1673-9418.2103101
收稿日期:
2021-03-29
修回日期:
2021-06-15
出版日期:
2022-08-01
发布日期:
2021-06-09
通讯作者:
+E-mail: lyandcxh@nuaa.edu.cn作者简介:
潘玉(1997—),女,安徽阜阳人,硕士研究生,主要研究方向为模式识别、人工智能。基金资助:
PAN Yu1, CHEN Xiaohong+(), LI Shunming2, LI Jiyong3
Received:
2021-03-29
Revised:
2021-06-15
Online:
2022-08-01
Published:
2021-06-09
About author:
PAN Yu, born in 1997, M.S. candidate. Her res-earch interests include pattern recognition and artificial intelligence.Supported by:
摘要:
增量学习是处理大规模动态流数据的重要技术,在机器学习领域得到广泛应用。已有众多学者将其与降维方法相结合得到增量式降维算法,其中增量典型相关分析(ICCA)是典型相关分析(CCA)的增量式改进版本,可有效处理多视图的高维数据流降维问题。由于ICCA每次只利用单对样本更新投影向量,每新增一对样本均需更新一次投影向量,导致该算法比较耗时。为了提高算法的效率,提出了块增量典型相关分析(CICCA)算法。该算法无需计算样本协方差矩阵,直接将数据流按批处理,每次利用新增的批样本信息对上一步投影向量进行修正更新,从而得到主投影向量。进一步,在投影向量的正交补空间中计算其他投影向量,进而将原始高维的多视图数据投影到低维空间。在人工数据集和真实数据集上的实验结果表明,该算法提取低维特征的分类性能与CCA、ICCA相当,但训练时间大幅度减少。
中图分类号:
潘玉, 陈晓红, 李舜酩, 李纪永. 块增量典型相关分析[J]. 计算机科学与探索, 2022, 16(8): 1809-1818.
PAN Yu, CHEN Xiaohong, LI Shunming, LI Jiyong. Chunk Incremental Canonical Correlation Analysis[J]. Journal of Frontiers of Computer Science and Technology, 2022, 16(8): 1809-1818.
算法 | 时间复杂度 | 空间复杂度 |
---|---|---|
CCIPCA | ||
CCA | ||
ICCA | ||
CICCA |
表1 算法的复杂度
Table 1 Complexity of algorithm
算法 | 时间复杂度 | 空间复杂度 |
---|---|---|
CCIPCA | ||
CCA | ||
ICCA | ||
CICCA |
组合方式 | CCA | ICCA | CICCA |
---|---|---|---|
串行 | 93.45(0.24) | 93.76(0.23) | 93.19(0.16) |
并行 | 93.57(0.17) | 93.48(0.18) | 93.75(0.14) |
表2 人工数据集的分类准确率(I)
Table 2 Classification accuracy of synthetic dataset (I) %
组合方式 | CCA | ICCA | CICCA |
---|---|---|---|
串行 | 93.45(0.24) | 93.76(0.23) | 93.19(0.16) |
并行 | 93.57(0.17) | 93.48(0.18) | 93.75(0.14) |
组合方式 | CCA | ICCA | CICCA |
---|---|---|---|
串行 | 96.95(0.01) | 96.10(0.06) | 96.19(0.04) |
并行 | 95.71(0.12) | 95.67(0.04) | 95.86(0.04) |
表3 WebKB的分类准确率(I)
Table 3 Classification accuracy of WebKB (I) %
组合方式 | CCA | ICCA | CICCA |
---|---|---|---|
串行 | 96.95(0.01) | 96.10(0.06) | 96.19(0.04) |
并行 | 95.71(0.12) | 95.67(0.04) | 95.86(0.04) |
视图 | CCA | ICCA | CICCA | ||||
---|---|---|---|---|---|---|---|
串行 | 并行 | 串行 | 并行 | 串行 | 并行 | ||
cap | alt | 91.08(0.02) | 88.86(0.29) | 89.78(0.35) | 88.56(0.35) | 89.83(0.28) | 89.09(0.06) |
cap | url | 91.99(0.07) | 91.46(0.03) | 91.61(0.09) | 91.99(0.09) | 92.07(0.03) | 90.31(0.04) |
cap | orig | 90.47(0.15) | 89.40(0.16) | 90.24(0.16) | 88.41(0.43) | 90.92(0.01) | 88.48(0.22) |
cap | ancurl | 94.58(0.01) | 94.13(0.02) | 93.06(0.09) | 92.30(0.08) | 93.82(0.02) | 93.90(0.09) |
alt | url | 89.32(0.34) | 87.72(0.21) | 91.53(0.19) | 88.48(0.25) | 88.02(0.35) | 87.57(0.34) |
alt | orig | 87.11(0.40) | 87.95(0.05) | 88.18(0.42) | 85.43(0.13) | 87.64(0.23) | 88.86(0.16) |
alt | ancurl | 91.84(0.17) | 89.17(0.11) | 93.90(0.01) | 90.54(0.10) | 93.82(0.09) | 91.63(0.02) |
url | orig | 90.54(0.22) | 91.41(0.18) | 90.24(0.16) | 91.56(0.26) | 92.60(0.02) | 93.75(0.05) |
url | ancurl | 92.37(0.25) | 90.71(0.09) | 91.53(0.10) | 90.40(0.43) | 93.99(0.03) | 93.82(0.11) |
orig | ancurl | 91.99(0.34) | 92.02(0.24) | 92.83(0.13) | 92.70(0.36) | 92.91(0.16) | 93.67(0.19) |
Avg | 91.13(0.23) | 90.28(0.13) | 91.29(0.17) | 90.03(0.24) | 91.56(0.12) | 91.11(0.12) |
表4 Ads的分类准确率(I)
Table 4 Classification accuracy of Ads (I) %
视图 | CCA | ICCA | CICCA | ||||
---|---|---|---|---|---|---|---|
串行 | 并行 | 串行 | 并行 | 串行 | 并行 | ||
cap | alt | 91.08(0.02) | 88.86(0.29) | 89.78(0.35) | 88.56(0.35) | 89.83(0.28) | 89.09(0.06) |
cap | url | 91.99(0.07) | 91.46(0.03) | 91.61(0.09) | 91.99(0.09) | 92.07(0.03) | 90.31(0.04) |
cap | orig | 90.47(0.15) | 89.40(0.16) | 90.24(0.16) | 88.41(0.43) | 90.92(0.01) | 88.48(0.22) |
cap | ancurl | 94.58(0.01) | 94.13(0.02) | 93.06(0.09) | 92.30(0.08) | 93.82(0.02) | 93.90(0.09) |
alt | url | 89.32(0.34) | 87.72(0.21) | 91.53(0.19) | 88.48(0.25) | 88.02(0.35) | 87.57(0.34) |
alt | orig | 87.11(0.40) | 87.95(0.05) | 88.18(0.42) | 85.43(0.13) | 87.64(0.23) | 88.86(0.16) |
alt | ancurl | 91.84(0.17) | 89.17(0.11) | 93.90(0.01) | 90.54(0.10) | 93.82(0.09) | 91.63(0.02) |
url | orig | 90.54(0.22) | 91.41(0.18) | 90.24(0.16) | 91.56(0.26) | 92.60(0.02) | 93.75(0.05) |
url | ancurl | 92.37(0.25) | 90.71(0.09) | 91.53(0.10) | 90.40(0.43) | 93.99(0.03) | 93.82(0.11) |
orig | ancurl | 91.99(0.34) | 92.02(0.24) | 92.83(0.13) | 92.70(0.36) | 92.91(0.16) | 93.67(0.19) |
Avg | 91.13(0.23) | 90.28(0.13) | 91.29(0.17) | 90.03(0.24) | 91.56(0.12) | 91.11(0.12) |
视图 | CCA | ICCA | CICCA | ||||
---|---|---|---|---|---|---|---|
串行 | 并行 | 串行 | 并行 | 串行 | 并行 | ||
fac | fou | 93.27(0.06) | 85.58(0.11) | 93.55(0.09) | 83.51(0.09) | 85.12(0.09) | 84.83(0.15) |
fac | kar | 92.71(0.04) | 85.65(0.37) | 92.34(0.03) | 84.73(0.17) | 88.96(0.15) | 84.82(0.04) |
fac | mor | 89.06(0.11) | 89.70(0.12) | 91.61(0.08) | 90.95(0.09) | 85.96(0.18) | 90.98(0.08) |
fac | pix | 91.21(0.01) | 81.91(0.17) | 93.29(0.01) | 87.23(0.05) | 93.31(0.01) | 87.25(0.05) |
fac | zer | 91.28(0.12) | 91.42(0.04) | 88.51(0.29) | 83.67(0.13) | 83.07(0.24) | 83.75(0.18) |
fou | kar | 93.80(0.09) | 89.98(0.04) | 91.77(0.05) | 85.51(0.16) | 93.12(0.08) | 88.25(0.09) |
fou | mor | 90.70(0.03) | 91.78(0.16) | 91.12(0.09) | 91.17(0.13) | 91.97(0.11) | 92.60(0.12) |
fou | pix | 91.98(0.04) | 91.58(0.32) | 93.32(0.06) | 92.10(0.09) | 93.40(0.26) | 92.80(0.15) |
fou | zer | 92.94(0.15) | 90.85(0.15) | 91.72(0.35) | 90.40(0.03) | 91.84(0.22) | 91.43(0.09) |
kar | mor | 92.06(0.01) | 92.33(0.10) | 94.28(0.09) | 94.16(0.16) | 94.32(0.06) | 92.98(0.05) |
kar | pix | 94.59(0.12) | 94.42(0.09) | 95.09(0.01) | 94.59(0.11) | 95.60(0.13) | 95.59(0.33) |
kar | zer | 91.62(0.18) | 91.81(0.18) | 91.00(0.32) | 91.30(0.21) | 91.09(0.24) | 91.10(0.20) |
mor | pix | 89.63(0.09) | 90.01(0.16) | 90.62(0.09) | 92.25(0.25) | 90.47(0.07) | 92.40(0.09) |
mor | zer | 90.49(0.19) | 88.14(0.13) | 83.70(0.46) | 87.36(0.17) | 84.86(0.35) | 88.87(0.14) |
pix | zer | 90.25(0.12) | 90.72(0.08) | 90.88(0.11) | 90.84(0.02) | 90.91(0.18) | 90.83(0.04) |
Avg | 91.71(0.11) | 89.73(0.18) | 91.52(0.20) | 89.32(0.15) | 90.27(0.19) | 89.90(0.16) |
表5 MFD的分类准确率(I)
Table 5 Classification accuracy of MFD (I) %
视图 | CCA | ICCA | CICCA | ||||
---|---|---|---|---|---|---|---|
串行 | 并行 | 串行 | 并行 | 串行 | 并行 | ||
fac | fou | 93.27(0.06) | 85.58(0.11) | 93.55(0.09) | 83.51(0.09) | 85.12(0.09) | 84.83(0.15) |
fac | kar | 92.71(0.04) | 85.65(0.37) | 92.34(0.03) | 84.73(0.17) | 88.96(0.15) | 84.82(0.04) |
fac | mor | 89.06(0.11) | 89.70(0.12) | 91.61(0.08) | 90.95(0.09) | 85.96(0.18) | 90.98(0.08) |
fac | pix | 91.21(0.01) | 81.91(0.17) | 93.29(0.01) | 87.23(0.05) | 93.31(0.01) | 87.25(0.05) |
fac | zer | 91.28(0.12) | 91.42(0.04) | 88.51(0.29) | 83.67(0.13) | 83.07(0.24) | 83.75(0.18) |
fou | kar | 93.80(0.09) | 89.98(0.04) | 91.77(0.05) | 85.51(0.16) | 93.12(0.08) | 88.25(0.09) |
fou | mor | 90.70(0.03) | 91.78(0.16) | 91.12(0.09) | 91.17(0.13) | 91.97(0.11) | 92.60(0.12) |
fou | pix | 91.98(0.04) | 91.58(0.32) | 93.32(0.06) | 92.10(0.09) | 93.40(0.26) | 92.80(0.15) |
fou | zer | 92.94(0.15) | 90.85(0.15) | 91.72(0.35) | 90.40(0.03) | 91.84(0.22) | 91.43(0.09) |
kar | mor | 92.06(0.01) | 92.33(0.10) | 94.28(0.09) | 94.16(0.16) | 94.32(0.06) | 92.98(0.05) |
kar | pix | 94.59(0.12) | 94.42(0.09) | 95.09(0.01) | 94.59(0.11) | 95.60(0.13) | 95.59(0.33) |
kar | zer | 91.62(0.18) | 91.81(0.18) | 91.00(0.32) | 91.30(0.21) | 91.09(0.24) | 91.10(0.20) |
mor | pix | 89.63(0.09) | 90.01(0.16) | 90.62(0.09) | 92.25(0.25) | 90.47(0.07) | 92.40(0.09) |
mor | zer | 90.49(0.19) | 88.14(0.13) | 83.70(0.46) | 87.36(0.17) | 84.86(0.35) | 88.87(0.14) |
pix | zer | 90.25(0.12) | 90.72(0.08) | 90.88(0.11) | 90.84(0.02) | 90.91(0.18) | 90.83(0.04) |
Avg | 91.71(0.11) | 89.73(0.18) | 91.52(0.20) | 89.32(0.15) | 90.27(0.19) | 89.90(0.16) |
视图 | CCIPCA | CCA | ICCA | CICCA |
---|---|---|---|---|
91.37(0.23) | 92.45(0.18) | 91.03(0.16) | 92.48(0.11) | |
91.37(0.23) | 92.32(0.01) | 91.75(0.32) | 92.61(0.10) |
表6 人工数据集的分类准确率(II)
Table 6 Classification accuracy of synthetic dataset (II) %
视图 | CCIPCA | CCA | ICCA | CICCA |
---|---|---|---|---|
91.37(0.23) | 92.45(0.18) | 91.03(0.16) | 92.48(0.11) | |
91.37(0.23) | 92.32(0.01) | 91.75(0.32) | 92.61(0.10) |
视图 | CCIPCA | CCA | ICCA | CICCA |
---|---|---|---|---|
79.40(0.12) | 83.61(0.12) | 81.55(0.11) | 81.33(0.10) | |
80.21(0.09) | 82.16(0.12) | 82.11(0.11) | 82.38(0.12) |
表7 WebKB的分类准确率(II)
Table 7 Classification accuracy of WebKB (II) %
视图 | CCIPCA | CCA | ICCA | CICCA |
---|---|---|---|---|
79.40(0.12) | 83.61(0.12) | 81.55(0.11) | 81.33(0.10) | |
80.21(0.09) | 82.16(0.12) | 82.11(0.11) | 82.38(0.12) |
视图 | CCIPCA | CCA | ICCA | CICCA | |||||
---|---|---|---|---|---|---|---|---|---|
cap | alt | 86.87(0.12) | 83.66(0.05) | 86.59(0.05) | 88.60(0.05) | 86.63(0.05) | 89.13(0.03) | 86.66(0.05) | 88.85(0.04) |
cap | url | 86.87(0.12) | 91.62(0.17) | 92.12(0.01) | 86.45(0.03) | 92.28(0.05) | 86.32(0.03) | 92.83(0.03) | 86.38(0.04) |
cap | orig | 86.87(0.12) | 89.37(0.10) | 86.74(0.16) | 86.61(0.14) | 86.86(0.01) | 86.58(0.04) | 87.06(0.04) | 86.83(0.13) |
cap | ancurl | 86.87(0.12) | 94.90(0.14) | 93.77(0.01) | 89.77(0.01) | 93.71(0.02) | 89.64(0.12) | 93.63(0.01) | 89.89(0.01) |
alt | url | 83.66(0.05) | 91.62(0.17) | 91.35(0.18) | 87.59(0.12) | 91.88(0.17) | 88.88(0.08) | 92.78(0.02) | 87.65(0.17) |
alt | orig | 83.66(0.05) | 89.37(0.10) | 86.57(0.04) | 81.86(0.02) | 86.48(0.04) | 83.16(0.02) | 87.43(0.01) | 81.73(0.02) |
alt | ancurl | 83.66(0.05) | 94.90(0.14) | 82.74(0.02) | 91.04(0.07) | 83.45(0.13) | 93.77(0.02) | 82.49(0.01) | 93.17(0.03) |
url | orig | 91.62(0.17) | 89.37(0.10) | 90.72(0.13) | 91.44(0.04) | 91.24(0.17) | 90.67(0.03) | 91.91(0.15) | 90.54(0.04) |
url | ancurl | 91.62(0.17) | 94.90(0.14) | 89.60(0.02) | 91.53(0.07) | 90.21(0.06) | 93.73(0.03) | 89.51(0.08) | 92.03(0.10) |
orig | ancurl | 89.37(0.10) | 94.90(0.14) | 90.09(0.11) | 91.01(0.14) | 89.89(0.15) | 93.10(0.12) | 90.37(0.14) | 91.69(0.12) |
Avg | — | — | 89.03(0.07) | 88.59(0.07) | 89.26(0.08) | 89.50(0.05) | 89.47(0.05) | 88.87(0.07) |
表8 Ads的分类准确率(II)
Table 8 Classification accuracy of Ads (II) %
视图 | CCIPCA | CCA | ICCA | CICCA | |||||
---|---|---|---|---|---|---|---|---|---|
cap | alt | 86.87(0.12) | 83.66(0.05) | 86.59(0.05) | 88.60(0.05) | 86.63(0.05) | 89.13(0.03) | 86.66(0.05) | 88.85(0.04) |
cap | url | 86.87(0.12) | 91.62(0.17) | 92.12(0.01) | 86.45(0.03) | 92.28(0.05) | 86.32(0.03) | 92.83(0.03) | 86.38(0.04) |
cap | orig | 86.87(0.12) | 89.37(0.10) | 86.74(0.16) | 86.61(0.14) | 86.86(0.01) | 86.58(0.04) | 87.06(0.04) | 86.83(0.13) |
cap | ancurl | 86.87(0.12) | 94.90(0.14) | 93.77(0.01) | 89.77(0.01) | 93.71(0.02) | 89.64(0.12) | 93.63(0.01) | 89.89(0.01) |
alt | url | 83.66(0.05) | 91.62(0.17) | 91.35(0.18) | 87.59(0.12) | 91.88(0.17) | 88.88(0.08) | 92.78(0.02) | 87.65(0.17) |
alt | orig | 83.66(0.05) | 89.37(0.10) | 86.57(0.04) | 81.86(0.02) | 86.48(0.04) | 83.16(0.02) | 87.43(0.01) | 81.73(0.02) |
alt | ancurl | 83.66(0.05) | 94.90(0.14) | 82.74(0.02) | 91.04(0.07) | 83.45(0.13) | 93.77(0.02) | 82.49(0.01) | 93.17(0.03) |
url | orig | 91.62(0.17) | 89.37(0.10) | 90.72(0.13) | 91.44(0.04) | 91.24(0.17) | 90.67(0.03) | 91.91(0.15) | 90.54(0.04) |
url | ancurl | 91.62(0.17) | 94.90(0.14) | 89.60(0.02) | 91.53(0.07) | 90.21(0.06) | 93.73(0.03) | 89.51(0.08) | 92.03(0.10) |
orig | ancurl | 89.37(0.10) | 94.90(0.14) | 90.09(0.11) | 91.01(0.14) | 89.89(0.15) | 93.10(0.12) | 90.37(0.14) | 91.69(0.12) |
Avg | — | — | 89.03(0.07) | 88.59(0.07) | 89.26(0.08) | 89.50(0.05) | 89.47(0.05) | 88.87(0.07) |
视图 | CCIPCA | CCA | ICCA | CICCA | |||||
---|---|---|---|---|---|---|---|---|---|
fac | fou | 88.49(0.19) | 88.60(0.01) | 88.90(0.13) | 90.59(0.06) | 90.81(0.19) | 89.11(0.01) | 90.34(0.02) | 89.75(0.01) |
fac | kar | 88.49(0.19) | 90.63(0.02) | 86.90(0.10) | 85.30(0.17) | 88.71(0.25) | 88.36(0.17) | 88.72(0.14) | 88.72(0.14) |
fac | mor | 88.49(0.19) | 88.56(0.02) | 87.81(0.02) | 89.80(0.02) | 89.10(0.18) | 85.41(0.10) | 88.94(0.09) | 86.69(0.01) |
fac | pix | 88.49(0.19) | 91.50(0.04) | 91.38(0.01) | 93.89(0.09) | 91.44(0.01) | 93.61(0.01) | 91.55(0.04) | 94.85(0.02) |
fac | zer | 88.49(0.19) | 91.79(0.08) | 91.84(0.23) | 90.58(0.03) | 90.36(0.20) | 89.71(0.15) | 89.59(0.03) | 89.89(0.08) |
fou | kar | 88.60(0.01) | 90.63(0.02) | 87.10(0.01) | 93.56(0.01) | 87.33(0.12) | 89.91(0.01) | 85.69(0.29) | 89.10(0.03) |
fou | mor | 88.60(0.01) | 88.56(0.02) | 88.24(0.36) | 90.70(0.02) | 88.05(0.02) | 91.00(0.01) | 87.24(0.02) | 91.54(0.02) |
fou | pix | 88.60(0.01) | 91.50(0.04) | 85.36(0.02) | 88.67(0.01) | 86.91(0.02) | 83.54(0.09) | 85.88(0.01) | 83.85(0.09) |
fou | zer | 88.60(0.01) | 91.79(0.08) | 90.11(0.23) | 86.75(0.13) | 89.65(0.15) | 86.64(0.27) | 89.10(0.17) | 86.84(0.23) |
kar | mor | 90.63(0.02) | 88.56(0.02) | 94.66(0.01) | 94.15(0.02) | 94.02(0.01) | 94.50(0.01) | 94.78(0.01) | 94.66(0.01) |
kar | pix | 90.63(0.02) | 91.50(0.04) | 94.15(0.01) | 94.23(0.04) | 95.71(0.01) | 95.93(0.09) | 94.76(0.14) | 93.08(0.12) |
kar | zer | 90.63(0.02) | 91.79(0.08) | 86.06(0.18) | 87.01(0.24) | 90.53(0.21) | 87.53(0.10) | 89.25(0.04) | 87.11(0.09) |
mor | pix | 88.56(0.02) | 91.50(0.04) | 89.11(0.01) | 88.08(0.32) | 89.76(0.11) | 89.54(0.01) | 89.92(0.02) | 89.65(0.01) |
mor | zer | 88.56(0.02) | 91.79(0.08) | 89.11(0.06) | 89.69(0.14) | 86.76(0.13) | 89.73(0.07) | 85.92(0.12) | 89.91(0.13) |
pix | zer | 91.50(0.04) | 91.79(0.08) | 90.11(0.13) | 90.23(0.01) | 90.52(0.36) | 89.49(0.17) | 90.70(0.01) | 90.02(0.22) |
Avg | — | — | 89.39(0.10) | 90.22(0.08) | 89.98(0.13) | 89.61(0.09) | 89.49(0.08) | 89.71(0.01) |
表9 MFD的分类准确率(II)
Table 9 Classification accuracy of MFD (II) %
视图 | CCIPCA | CCA | ICCA | CICCA | |||||
---|---|---|---|---|---|---|---|---|---|
fac | fou | 88.49(0.19) | 88.60(0.01) | 88.90(0.13) | 90.59(0.06) | 90.81(0.19) | 89.11(0.01) | 90.34(0.02) | 89.75(0.01) |
fac | kar | 88.49(0.19) | 90.63(0.02) | 86.90(0.10) | 85.30(0.17) | 88.71(0.25) | 88.36(0.17) | 88.72(0.14) | 88.72(0.14) |
fac | mor | 88.49(0.19) | 88.56(0.02) | 87.81(0.02) | 89.80(0.02) | 89.10(0.18) | 85.41(0.10) | 88.94(0.09) | 86.69(0.01) |
fac | pix | 88.49(0.19) | 91.50(0.04) | 91.38(0.01) | 93.89(0.09) | 91.44(0.01) | 93.61(0.01) | 91.55(0.04) | 94.85(0.02) |
fac | zer | 88.49(0.19) | 91.79(0.08) | 91.84(0.23) | 90.58(0.03) | 90.36(0.20) | 89.71(0.15) | 89.59(0.03) | 89.89(0.08) |
fou | kar | 88.60(0.01) | 90.63(0.02) | 87.10(0.01) | 93.56(0.01) | 87.33(0.12) | 89.91(0.01) | 85.69(0.29) | 89.10(0.03) |
fou | mor | 88.60(0.01) | 88.56(0.02) | 88.24(0.36) | 90.70(0.02) | 88.05(0.02) | 91.00(0.01) | 87.24(0.02) | 91.54(0.02) |
fou | pix | 88.60(0.01) | 91.50(0.04) | 85.36(0.02) | 88.67(0.01) | 86.91(0.02) | 83.54(0.09) | 85.88(0.01) | 83.85(0.09) |
fou | zer | 88.60(0.01) | 91.79(0.08) | 90.11(0.23) | 86.75(0.13) | 89.65(0.15) | 86.64(0.27) | 89.10(0.17) | 86.84(0.23) |
kar | mor | 90.63(0.02) | 88.56(0.02) | 94.66(0.01) | 94.15(0.02) | 94.02(0.01) | 94.50(0.01) | 94.78(0.01) | 94.66(0.01) |
kar | pix | 90.63(0.02) | 91.50(0.04) | 94.15(0.01) | 94.23(0.04) | 95.71(0.01) | 95.93(0.09) | 94.76(0.14) | 93.08(0.12) |
kar | zer | 90.63(0.02) | 91.79(0.08) | 86.06(0.18) | 87.01(0.24) | 90.53(0.21) | 87.53(0.10) | 89.25(0.04) | 87.11(0.09) |
mor | pix | 88.56(0.02) | 91.50(0.04) | 89.11(0.01) | 88.08(0.32) | 89.76(0.11) | 89.54(0.01) | 89.92(0.02) | 89.65(0.01) |
mor | zer | 88.56(0.02) | 91.79(0.08) | 89.11(0.06) | 89.69(0.14) | 86.76(0.13) | 89.73(0.07) | 85.92(0.12) | 89.91(0.13) |
pix | zer | 91.50(0.04) | 91.79(0.08) | 90.11(0.13) | 90.23(0.01) | 90.52(0.36) | 89.49(0.17) | 90.70(0.01) | 90.02(0.22) |
Avg | — | — | 89.39(0.10) | 90.22(0.08) | 89.98(0.13) | 89.61(0.09) | 89.49(0.08) | 89.71(0.01) |
[1] | HOTELLING H. Analysis of a complex of statistical varia-bles into principal components[J]. Journal of Educational Psychology, 1933, 24: 417-441. |
[2] | FISHER R A. The use of multiple measurements in taxo-nomic problems[J]. Annals of Eugenics, 1936, 7: 179-188. |
[3] | KOWALSKI G B R. Partial least-squares regression: a tuto-rial[J]. Analytica Chimica Acta, 1986, 185: 1-17. |
[4] | HOTELLING H. Relations between two sets of variates[J]. Biometrika, 1935, 28: 321-377. |
[5] | 曹杰, 顾斌杰, 熊伟丽. 增量式约简最小二乘孪生支持向量回归机[J]. 计算机科学与探索, 2021, 15(3): 553-563. |
CAO J, GU B J, XIONG W L. Incremental reduced least squares twin support vector regression[J]. Journal of Fron-tiers of Computer Science and Technology, 2021, 15(3): 553-563. | |
[6] | WENG J Y, ZHANG Y L, WANG W S. Candid covariance-free incremental principal component analysis[J]. IEEE Tran-sactions on Pattern Analysis and Machine Intelligence, 2003, 25(8): 1034-1040. |
[7] | GE W, SUN M, WANG X. An incremental two-dimensional principal component analysis for object recognition[J]. Mat-hematical Problems in Engineering, 2018(3): 1-13. |
[8] | STUTTAFORD S A, KRASOULIS A, DUPAN S S G. Auto-matic myoelectric control site detection using candid cova-riance-free incremental principal component analysis[C]// Pro-ceedings of the 42nd Annual International Conference of the IEEE Engineering in Medicine & Biology Society, Mont-real, Jul 20-24, 2020. Piscataway: IEEE, 2020: 3497-3500. |
[9] | MISHIN D, BRANTNER M K, CZAKO F, et al. Real time change point detection by incremental PCA in large scale sensor data[C]// Proceedings of the 2014 IEEE High Perfor-mance Extreme Computing Conference, Waltham, Sep 9-11, 2014. Piscataway: IEEE, 2014: 1-6. |
[10] | FUJIWARA T, CHOU J K. An incremental dimensionality reduction method for visualizing streaming multi-dimensional data[J]. IEEE Transactions on Visualization and Computer Graphics, 2020, 26(1): 418-428. |
[11] | CHU D L, LIAO L Z, WANG X Y. Incremental linear disc-riminant analysis: a fast algorithm and comparisons[J]. IEEE Transactions on Neural Networks and Learning Systems, 2017, 26(11): 2716-2735. |
[12] | ZENG X Q, LI G Z. Dimension reduction for p53 protein recognition by using incremental partial least squares[C]// Proceedings of the 2013 IEEE International Conference on Bioinformatics and Biomedicine, Shanghai, Dec 18-21, 2013. Washington: IEEE Computer Society, 2013: 381-385. |
[13] | SIMONCINI V. Analysis of the rational Krylov subspace pro-jection method for large-scale algebraic Riccati equations[J]. SIAM Journal on Matrix Analysis and Applications, 2016, 37(4): 1655-1674. |
[14] | 牟昭曦, 陈晓红. 增量典型相关分析[D]. 南京: 南京航空航天大学, 2019. |
MOU Z X, CHEN X H. Incremental canonical correlation analysis[D]. Nanjing: Nanjing University of Aeronautics and Astronautics, 2019. | |
[15] | OZAWA S, HIRAI M, ABE S. An online face recognition system with incremental learning ability[C]// Proceedings of the SICE Annual Conference 2007, Takamatsu, Sep 17-20, 2007. Piscataway: IEEE, 2007: 1968-1971. |
[16] | PANG S, OZAWA S, KASABOV N. Incremental linear discriminant analysis for classification of data streams[J]. IEEE Transactions on Systems, Man, and Cybernetics, Part B: Cybernetics, 2005, 35(5): 905-914. |
[17] | 曾雪强, 叶震麟, 左家莉. 一种块增量偏最小二乘算法[J]. 山东大学学报(理学版), 2019, 54(3): 93-101. |
ZENG X Q, YE Z L, ZUO J L. A chunk increment partial least squares algorithm[J]. Journal of Shandong University (Natural Science), 2019, 54(3): 93-101. | |
[18] | CLAYPO N, HANSKUNATAI A, JAIYEN S. A new strea-ming learning for stream chunk data classification based on incremental learning and adaptive Boosting algorithm[C]// Proceedings of the 13th International Conference on Com-puting and Information Technology, Bangkok, Jul 6-7, 2017. Cham: Springer, 2017: 160-168. |
[19] | 李航. 统计学习方法[M]. 北京: 清华大学出版社, 2012. |
LI H. Statistical learning method[M]. Beijing: Tsinghua University Press, 2012. | |
[20] | 刘季, 陈秀宏, 杭文龙. 面向大规模数据的快速多代表点仿射传播算法[J]. 计算机科学与探索, 2016, 10(2): 268-276. |
LIU J, CHEN X H, HANG W L. Fast multi-representative affine propagation algorithm for large scale data[J]. Journal of Frontiers of Computer Science and Technology, 2016, 10(2): 268-276. | |
[21] | 何云斌, 董恒, 万静. 基于密度峰值和近邻优化的聚类算法[J]. 计算机科学与探索, 2020, 14(4): 554-565. |
HE Y B, DONG H, WAN J. Clustering algorithm based on density peak and neighbor optimization[J]. Journal of Frontiers of Computer Science and Technology, 2020, 14(4): 554-565. |
[1] | 夏笑秋, 陈松灿. 改进的二视图随机森林[J]. 计算机科学与探索, 2022, 16(1): 144-152. |
[2] | 杨悦, 王士同. 随机特征映射的四层神经网络及其增量学习[J]. 计算机科学与探索, 2021, 15(7): 1265-1278. |
[3] | 倪鹏,刘阳明,赵素云,陈红,李翠平. 动态模糊粗糙特征选取算法[J]. 计算机科学与探索, 2020, 14(2): 236-243. |
[4] | 何丽,韩克平,刘颖. 自适应的SVM增量算法[J]. 计算机科学与探索, 2019, 13(4): 647-656. |
[5] | 王宏杰,滕飞,李天瑞. 模块度引导下的社区发现增量学习算法[J]. 计算机科学与探索, 2017, 11(4): 556-564. |
[6] | 刘三民,王忠群,刘涛,修宇. 融合互近邻降噪的动态数据流分类研究[J]. 计算机科学与探索, 2016, 10(1): 36-42. |
[7] | 孙桃,谢振平,王士同,刘渊. 容量约束的自组织增量联想记忆模型[J]. 计算机科学与探索, 2016, 10(1): 130-141. |
[8] | 丁 鑫,陈晓红,陈松灿. 核诱导距离度量的鲁棒典型相关分析[J]. 计算机科学与探索, 2012, 6(8): 708-716. |
阅读次数 | ||||||
全文 |
|
|||||
摘要 |
|
|||||