Differences
This shows you the differences between two versions of the page.
Both sides previous revisionPrevious revisionNext revision | Previous revision | ||
data_mining:hmm [2014/11/22 02:13] – [Parameter] phreazer | data_mining:hmm [2014/12/17 00:47] (current) – [Beispiel] phreazer | ||
---|---|---|---|
Line 29: | Line 29: | ||
$T(i,j) = P(Z_{k+1}=j|z_k=i)$ ($i,j \in \{i, | $T(i,j) = P(Z_{k+1}=j|z_k=i)$ ($i,j \in \{i, | ||
- | Emission probabilities | + | T ist die Transition Matrix (Übergangswkt.) |
+ | |||
+ | Emission probabilities: | ||
$\varepsilon_i(x) = p(x|Z_k=i)$ für $i\in \{i, | $\varepsilon_i(x) = p(x|Z_k=i)$ für $i\in \{i, | ||
Line 39: | Line 41: | ||
pmf | pmf | ||
- | Initial distribution: | + | Initial distribution: |
+ | |||
+ | $\pi(i) = P(Z_i=i), \in \{i, | ||
Joint Distribution: | Joint Distribution: | ||
- | $p(x_1,...,x_n,z_1,...z_n) = \pi(z_1) \varepsilon_z_1(x_1) \prod_{k=2}^n T(z_{k-1}, | + | $p(x_1,\dots,x_n,z_1,\dots,z_n) = \pi(z_1) \varepsilon_{z_1}(x_1) \prod_{k=2}^n T(z_{k-1}, |
+ | |||
+ | |||
+ | |||
+ | ===== Forward-Backward Algorithmus ===== | ||
+ | |||
+ | ===== Beispiel ===== | ||
+ | Zustand ist Durschnittstemperatur: | ||
+ | Beobachtbarer Zustand ist Dicke der Ringe: S, M, L | ||
+ | |||
+ | Nun wird Abfolge S,M,S,L beobachtet. Daraus soll die wahrscheinlichste Zustandssequenz des Markovprozess ermittelt werden. |