blend_deeptamer_AIF.dot

digraph Blended_Concept {
    rankdir=TB;
    fontsize = 10;
    fontname = "Helvetica";
    
    node[shape=box, style=rounded, fontname=Helvetica, fontsize=10];
    edge[arrowhead=vee, arrowsize=0.5, fontname=Helvetica, fontsize=10];
    
    subgraph cluster_LLM {
        label="LLM (Thoughts, Mental Models, or Algorithms)";
        style=dotted;
        color=gray;
        LLM -> RR [label="Translates mental models into actions"];
        R -> LLM [label="Provides feedback to update mental models"];
        
        subgraph cluster_TAMER {
            label="Deep TAMER Algorithm";
            style=dotted;
            color=gray;
            
            Agent -> Environment [label="Observes state and executes action"];
            Environment -> Agent [label="Provides feedback in the form of reward signals"];
            Agent -> Memory [label="Stores transitions for experience replay"];
            Memory -> Trainer [label="Samples mini-batch for SGD update"];
            Trainer -> Hk_1 [label="Updates policy using SGD update (3)"];
            Hk_1 -> Agent [label="Sends updated policy to agent"];
            
            RR -> Agent [style=dashed, label="Uses learned policy to interact with environment"];
        }
    }
    
    subgraph cluster_R {
        label="R (External Reality)";
        
        Environment -> R [style=dashed, label="Generates state observations and rewards"];
        
        RR -> Environment [style=dashed, label="Translates actions into changes in external reality"];
    }
}