AOS55
commented
4 years ago Also getting this problem, stays highlighted after math mode. For example after second inline math $ all is highlighted in $
Conveniently $\frac{\nabla_{\bm{\theta}} \pi(A_{t}|S_{t}; \bm{\theta})}{\pi(A_{t}|S_{t}; \bm{\theta})}$ from equation\ref{eq:deriveScore} is the score function $\nabla_{\bm{\theta}} \ln(\pi(A_{t}|S_{t}; \bm{\theta}))$. This forms the basis of the REINFORCE algorithm \cite{RL41} and the update step samples $G_{t} \nabla_{\bm{\theta}} \ln(\pi(A_{t}|S_{t}; \bm{\theta}))$. If the return $G_{t}$ increases then the weight of the parameter is increased, therefore weighting policys with higher return. But reduces the size of the update if the probability of selecting an action is high, discounting policies that suffer from maximisation bias. REINFORCE is shown in algorithm \ref{algo:T-REINFORCE}, and best thought of as a parameterised MC method because the return is calculated based on the discounted sum for the whole episode.
bbgra
Prerequisites
Xin this box if you have done the following:Description
Steps to reproduce & Minimum working example
1.
In latex source, create a paragraph of: Isdadsa sdadas \cite{ds231} sadasds Fig.~\ref{fig:sss} by \cite{sdads12} dasda \SI{23}{\celsius}, dsad \SI{22222}{\celsius} adasd sad asd asd asd sa \SI{22222}{\celsius} dsadsa dnd \ce{Mg^2+} sds. \cite{sd223aa} sd \SI{21323123232}{\celsius} dsad \gls{sss} dasda \SI{2123322}{\celsius}.
The last \SI is not highlighted. If you delete/shorten the first cite to \cite{31}, it becomes highlighted.