Download presentation
Presentation is loading. Please wait.
Published byErik Kristian O’Brien’ Modified over 9 years ago
1
1 Information Geometry of Self-organizing maximum likelihood Shinto Eguchi ISM, GUAS This talk is based on joint research with Dr Yutaka Kano, Osaka Univ Bernoulli 2000 Conference at Riken on 27 October, 2000
2
2 Consider a statistical model: -MLE Maximum Likelihood Estimation (MLE)( Fisher, 1922), Take an increasing function. Consistency, efficiency sufficiency, unbiasedness invariance, information
3
3 -MLE MLE Normal density -MLE given data
4
4 -3-212 3 0.1 0.2 0.3 0.4 outlier MLE -MLE Normal density
5
5
6
6 Examples (1 ) (2 ) (3 ) -divergence KL-divergence
7
7
8
8 Pythagorian theorem (1,1) (1,0) (0,1) ( t, s ). (0,0) f g h
9
9 (Pf)
10
10 Differential geometry of Riemann metric Affine connection Conjugate affine connection Ciszsar’s divergence
11
11 -divergence Amari’s -divergence
12
12 -likelihood function M-estimation ( Huber, 1964, 1983) Kullback-Leibler and maximum likelihood
13
13 Another definition of -likelihood Take a positive function (x, ) and define -likelihood equation is a weighted score with integrabity.
14
14 Consistency of -MLE
15
15 Influence function Fisher consistency -contamination model of Asymptotic efficiency Robustness or Efficiency
16
16 Generalized linear model Regression model Estimating equation
17
17 Bernoulli regression Logistic regression
18
18 Misclassification model MLE
19
19 Group II Group II from Group I = from Logistic Discrimination Mislabel Group I 5 Group II 35 Group I
20
20 Misclassification Group I 5 data Group II 35 data
21
21 Poisson regression -likelihood function Canonical link -contamination model
22
22 Neural network
23
23 Input Output
24
24 -maximum likelihood Maximum likelihood
25
25 Classical procedure for PCA Self-organizing procedure Let off-line data.
26
26
27
27 Classic procedure Self-organizing procedure
28
28 Independent Component Analysis (Minami & Eguchi, 2000) F F
29
29 S S F Theorem (Semiparametric consistency) (Pf)
30
30 -likelihood satisfies the semiparametric consistency
31
31
32
32 Usual methodself-organizing method Blue dots Blue & red dots
33
33 150 the exponential power 50 http://www.ai.mit.edu/people/fisher/ica_data/
34
34 Concluding remark Bias potential function ? ! -Regression analysis -Discriminant analysis -PCA -ICA -sufficiency -factoriziable -exponential family -EM algorithm
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.