Web Science & Technologies University of Koblenz ▪ Landau, Germany Data Mining & Machine Learning Dipl.-Inf. Christoph Carl Kling
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Web Science & Technologies
University of Koblenz ▪ Landau, Germany
Data Mining & Machine Learning
Dipl.-Inf. Christoph Carl Kling
C. C. Kling NetHDP3 of 17
WeST
Probability Theoryn = 1 n >= 1
Bernoulli = Binomial for n = 1 Binomial
k = 2
k > 2
Multinomial
100
1
Multinomial for n = 1
p
n → ∞
Gaussian
MulivariateGaussian
1 2 3 k
p
number of successes
C. C. Kling NetHDP4 of 17
WeST
Experiment
Observations c (our Data)Hidden (latent) parameter p
Example: tossing a coin: 2 x head, 0 x tail
tail head
C. C. Kling NetHDP5 of 17
WeST
Latent Dirichlet Allocation
C. C. Kling NetHDP6 of 17
WeST
Parameter Estimation
Maximum likelihood estimation (MLE)
p = 1.0 !
C. C. Kling NetHDP7 of 17
WeST
Parameter Estimation
p = 1.0
C. C. Kling NetHDP8 of 17
WeST
Probabilistic models
p more likely is close to 0.5!
Prior probability
C. C. Kling NetHDP9 of 17
WeST
Beta distribution
Density of
C. C. Kling NetHDP10 of 17
WeST
Beta distribution
Beta(100,100)
C. C. Kling NetHDP11 of 17
WeST
Beta distribution
Beta(10,10)
C. C. Kling NetHDP12 of 17
WeST
Beta distribution
Beta(1,1)
C. C. Kling NetHDP13 of 17
WeST
Beta distribution
Beta(0.1,0.1)
C. C. Kling NetHDP14 of 17
WeST
Beta distribution
Beta(0.01,0.01)
C. C. Kling NetHDP15 of 17
WeST
Parameter Estimation
Maximum a posteriori estimation (MAP)
Bayesian inference
C. C. Kling NetHDP16 of 17
WeST
Parameter Estimation
Maximum a posteriori estimation (MAP)
Bayesian inference
C. C. Kling NetHDP17 of 17
WeST
Lineare Regression
y = Größe x1 = Geschlecht x2 = Gewicht
168 1 65
172 0 80
164 1 52
187 0 120
194 0 90
Related Documents
