Backward View SARSA(λ)
A TD error at one cell propagates backward through eligibility traces to all recently visited cells.
▶ Step
Run
Reset
agent
trace strength (eligibility)
goal (+1)
Parameters
λ (trace decay):
0.90
α (learning rate):
0.10
γ (discount):
0.95
Current Step
step
0
state s
–
action a
–
reward r
–
TD error δ
–
episode return
0