《模式识别》课程教学资源（PPT讲稿）Learning with information of features

Learning with information of features 2009-06-05 panec

LOGO Learning with information of features 2009-06-05

http:/parnec.nuaa.edu.cn Contents Motivation Incorporating prior knowledge on features into learning (AISTATS OT Regularized learning with networks of features (NIPS08 Conclusion

Company name www.themegallery.com Contents Motivation Regularized learning with networks of features (NIPS’08) Incorporating prior knowledge on features into learning (AISTATS’07) Conclusion

http:/parnec.nuaa.edu.cn Motivation Given data X∈R×dx1 2d n min ∑。1(,f(x)+f IF prior information of samples Manifold structure information LAPSVM Transformation invariance VSVM. ISSVM Permutation invariance 丌-SVM Imbalance information SVM for imbalance distribution Cluster structure information Structure sⅤM

Company name www.themegallery.com Motivation 11 12 1 21 22 2 1 2 d d n n nd n d x x x x x x x x x                  Given data X∈R n×d n i=1 min ( , ( )) || || i i F  l y f x f +  + prior information of samples Manifold structure information LAPSVM Transformation invariance VSVM, ISSVM Permutation invariance π- SVM Imbalance information SVM for imbalance distribution Cluster structure information Structure SVM

http:/parnec.nuaa.edu.cn Motivation 12x 2d n2 Information in the sample spa ace (Space spanned by samples)

Company name www.themegallery.com Motivation 11 12 1 21 22 2 1 2 d d n n nd n d x x x x x x x x x                       Information in the sample space (space spanned by samples)

http:/parnec.nuaa.edu.cn Motivation 12 Prior information in the feature or attribute space (Space spanned by features)

Company name www.themegallery.com Motivation 11 12 1 21 22 2 1 2 d d n n nd n d x x x x x x x x x                Prior information in the feature or attribute space (space spanned by features)

http:/parnec.nuaa.edu.cn Motivation 12 min 2i(, f(x, ))+2 l lF prior information of features for better generalization

Company name www.themegallery.com Motivation n i=1 min ( , ( )) || || i i F  l y f x f +  + prior information of features for better generalization 11 12 1 21 22 2 1 2 d d n n nd n d x x x x x x x x x               

http:/parnec.nuaa.edu.cn Contents Motivation Incorporating prior knowledge on features into learning (AISTATS) Regularized learning with networks of features (7Fs78) Conclusion

Company name www.themegallery.com Contents Motivation Regularized learning with networks of features (NIPS’08) Incorporating prior knowledge on features into learning (AISTATS’07) Conclusion

http:/parnec.nuaa.edu.cn Incorporating prior knowledge on features into learning(AISTATS'o7) ●MotⅤ ation OKernel design by meta-features ● a toy example handwritten digit recognition aided by meta-features O Towards a theory of meta-features

Company name www.themegallery.com Incorporating prior knowledge on features into learning (AISTATS’07) ⚫ Motivation ⚫Kernel design by meta-features ⚫ A toy example ⚫ Handwritten digit recognition aided by meta-features ⚫ Towards a theory of meta-features

http:/parnec.nuaa.edu.cn Incorporating prior knowledge on features into learning (A/STATS'O7 ●MotⅤ ation KErnel design by meta-features ● A toy examp le O Handwritten digit recognition aided by meta-features O Towards a theory of meta-features

Company name www.themegallery.com Incorporating prior knowledge on features into learning (AISTATS’07) ⚫ Motivation ⚫Kernel design by meta-features ⚫ A toy example ⚫ Handwritten digit recognition aided by meta-features ⚫ Towards a theory of meta-features

http:/parnec.nuaa.edu.cn Incorporating prior knowledge on features into learning (A/STATS07 Image recognition task Feature: pixel(gray level) Coordinate(x,y)of pixel can be treated as Feature of features: meta-feature Feature with similar meta-feature, or more specifically, adjacent pixel should be assigned similar weights Propose a framework incorporating meta-features into learning

Company name www.themegallery.com Incorporating prior knowledge on features into learning (AISTATS’07) Image recognition task Feature : pixel (gray level) Coordinate (x,y) of pixel can be treated as Feature of features: meta-feature Propose a framework incorporating meta-features into learning Feature with similar meta-feature, or more specifically, adjacent pixel should be assigned similar weights