Algorithm 2. Server side algorithm.

Let $S$ and $T$ be the spatial and temporal resolution vector, respectively, given by: $S=\{S{R}_1,S{R}_2,S{R}_3,S{R}_{4,}S{R}_5\}$, $T=\{T{R}_1,T{R}_2,T{R}_3,T{R}_4,T{R}_5\}$. (1)$\mathrm{Initially}\mathrm{set}S{R}_1$ and $T{R}_1$ to a default value $\left(S_{\mathrm{QCIF}},T_d\right)$; (2)Continue: Read feedback message $\left(status\right)$ from the client; If $(\mathrm{Status}==Stabilized)$,      continue with the existing setup; Else if $\left(\mathrm{Status}==Progressive\right)$,      call A-LSTM(status), ##Increase Spatial/Temporal resolution; Else if $\left(\mathrm{Status}==Fluctuated\right)$,      find Switch_time (bit rate),    call A-LSTM(status), ##Update Spatial/Temporal resolution; Else if $\left(\mathrm{Status}==Degraded\right)$,      call A-LSTM(status), ##decrease Spatial/Temporal resolution; Else if $\left(\mathrm{Status}==Non-monotonic\right)$,      wait till next feedback message arrives; (3)Continue till connection is terminated; A. Function Find_Switch_Time (bit rate): Calculate time for quality switch $T_{switch}=TQ{S}_{k+1}-TQ{S}_k$:      #where $TQ{S}_{k+1}$ is the time instant at the end of $k^{th}$ served quality switching,      # request, and $TQ{S}_k$ is the present time instant attending previous quality,      # switching request; Set a timer when request for quality switch is received; Wait for the next feedback message; If $\left(T_{switch}>{\mathrm{T}}_{Fluctuation\_time}\right)$,      discard the request for quality switching and wait for next client      feedback message; Else,      serve the request for quality switching.