Click a state-action pair (s, a) to see which parameters are affected by a gradient update
Tabular: θ = Q-table
10 parameters for 10 Q-values
Parameter vector θ and ∇
θ
Q(s, a)
Resulting Q-values
Cosine FA: Q
θ
(s,a) = θ₁ + θ₂cos(π·d/3) + θ₃a
3 parameters for 10 Q-values
Parameter vector θ and ∇
θ
Q(s, a)
Resulting Q-values