this is for holding javascript data
jBillou edited Hidden Markov Models.tex
about 9 years ago
Commit id: f9b819f2aa6a3893621fe6e72f3de00a6f22f740
deletions | additions
diff --git a/Hidden Markov Models.tex b/Hidden Markov Models.tex
index 3fb77fd..aa2e9a4 100644
--- a/Hidden Markov Models.tex
+++ b/Hidden Markov Models.tex
...
The HMM for inferring the cell cycle phase $\phi$ is very similar to the one for the circadian phase. The only difference is that we didn't included a baseline state, as it wasn't needed. The full HMM for the is defined by:
\begin{align*}
\label{eq:cellcylce_HMM} \begin{equation} \label{eqn:cellcylce_HMM}
\textrm{d}\phi_t
= \frac{2\pi}{T_2}\textrm{d}t + \sigma_{\phi}\textrm{d}W_t
\end{equation}
\begin{equation} \label{eqn:cellcylce_HMM2}
\textrm{d}\alpha_t = -\gamma_{\alpha} (\alpha_t-\mu_{\alpha}) \textrm{d}t + \sigma_{\alpha}\textrm{d}W_t
\end{equation}
\begin{equation} \label{eqn:cellcylce_HMM3}
a_t = \exp(\alpha_t) w(\phi_t) +\xi
\end{equation}
%\begin{align*}
%\label{eq:cellcylce_HMM}
%\textrm{d}\phi_t =& \frac{2\pi}{T_2}\textrm{d}t + \sigma_{\phi}\textrm{d}W_t\\
\textrm{d}\alpha_t %\textrm{d}\alpha_t =& -\gamma_{\alpha} (\alpha_t-\mu_{\alpha}) \textrm{d}t + \sigma_{\alpha}\textrm{d}W_t\\
s_t %s_t =& \exp(\alpha_t) w(\phi_t) +\xi\\
\end{align*} %\end{align*}
We used a mean period $T_2$ of $22$h and a phase diffusion coefficient $\sigma_{\phi}$ of $0.15$ $\textrm{rad} \; h^{-1/2}$. For the amplitude $\alpha$ we used a timescale $\gamma_{\alpha}^{-1}$ of $30$h a zero mean value $\mu_{\alpha}$ and a diffusion coefficient $\sigma_{\alpha}$ of $0.035$. The variance of the noise $\xi$ was set to $0.1$. The data were quantile-normalized as in previous section.