//// TD learning
clear; clf; 
gamma=0.8; reward=[0 1 0 0.5 0]; pattern_vector=rand(10,5);
w=rand(1,10); V_mem=0; previous_state=5;

for episode=1:100; TDerror(episode)=0;
    for pattern=1:5
        V=w*pattern_vector(:,pattern);
        rhat=reward(previous_state)+gamma*V-V_mem;
        w=w+0.2*rhat*pattern_vector(:,previous_state)';
        TDerror(episode)=TDerror(episode)+abs(rhat);
        previous_state=pattern; V_mem=V;
    end
end

subplot(2,1,1);
 plot(TDerror);xlabel('Episode'); ylabel('TD error')
subplot(2,1,2);
 plot(w*pattern_vector);xlabel('Pattern'); ylabel('V(Pattern)')