algorithm.tex

\documentclass{article}
\usepackage{graphicx} % Required for inserting images

\title{Algorithm}
\author{SAMIK SARASWAT 20BRS1205}

\date{April 2023}



\usepackage{algorithm,algorithmic}
\begin{document}
 \begin{algorithm}
 \caption{TD3 Algorithm}
 \begin{algorithmic}[1]
 \renewcommand{\algorithmicrequire}{\textbf{Input:}}
 \renewcommand{\algorithmicensure}{\textbf{Output:}}
 \REQUIRE $S, A, R(s, a), T(s, a, s'), \gamma, R, N, M, T, D, \rho$
 \ENSURE  $\nu(s, \theta_\nu, Q1(s, a, \theta_{Q1}), Q2(s, a, \theta_{Q2})$
 
 %\\ \textit{Initialisation} :
 
  \STATE $\theta^1 \leftarrow random$
  \STATE $\theta^2 \leftarrow random$
  \STATE $\theta^{\nu} \leftarrow random$
  \STATE Critic network initialization:
  \STATE $Q1(s,a; \theta^1) \leftarrow \theta^1$
  \STATE $Q2(s,a; \theta^2) \leftarrow \theta^2$
  \STATE ${\nu}(s,a; \theta^{\nu}) \leftarrow \theta^{\nu}$
  \STATE Target network initialization:
  \STATE $Q1^{'}(s,a; \theta^1) \leftarrow \theta^1$
  \STATE $Q2^{'}(s,a; \theta^2) \leftarrow \theta^2$
  \STATE ${\nu}^{'}(s,a; \theta^{\nu}) \leftarrow \theta^{\nu}$
  
  \FOR {$i = 1 to M$}
    \FOR{$j = 1 to T$} 
      \STATE $a \leftarrow \nu(s) + N$
      \STATE $s', r = ENV-STEP(a)$
      \STATE $R \leftarrow (s, a, r, s')$
      \STATE $B \leftarrow SAMPLE(R)$
      \STATE $y \leftarrow r + \gamma*min(Q1'(s', \nu'(s')), Q2'(s',\nu'(s')))$
      \STATE $L \leftarrow (y - Q1(s, a)^2 + (y - Q2(s, a))^2$
      \STATE $\theta_{Q1} \leftarrow \theta_{Q1} - \alpha_Q*\nabla_{\theta_{Q1}}L$ 
      \STATE $\theta_{Q2} \leftarrow \theta_{Q2} - \alpha_Q*\nabla_{\theta_{Q2}}L$ 
      
      \IF {$i \% d = 0$}
          \STATE $\nu(s) = max( Q1(s,\nu(s)) + Q2(s, \nu(s))$
          \STATE $Q1^{'}(s,a) = \rho*Q1^{'}(s,a) + (1- \rho) * Q1(s,a)$
          \STATE $Q2^{'}(s,a) =\rho*Q2^{'}(s,a) + (1- \rho) * Q2(s,a)$
          \STATE $\nu^{'}(s) =\rho* \nu^{'}(s) + (1-\rho) * \nu(s)$
      \ENDIF
    \ENDFOR
  \ENDFOR
 \RETURN $P$ 
 \end{algorithmic} 
 \end{algorithm}

\end{document}