eps = 100 lstd = td.RecursiveLSTDLambda(lam=lam, eps=eps, phi=phi) lstd.name = r"LSTD({})".format(lam) methods.append(lstd) lam = .8 alpha = 1. lspe = td.RecursiveLSPELambda(lam=lam, alpha=alpha, phi=phi) lspe.name = r"LSPE({}) $\alpha$={}".format(lam, alpha) methods.append(lspe) lam = .0 alpha = .01 beta = 1000 mins = 0 lstd = td.FPKF(lam=lam, alpha=alpha, beta=beta, mins=mins, phi=phi) lstd.name = r"FPKF({}) $\alpha={}$ $\beta={}$".format(lam, alpha, beta) lstd.ls = "--" methods.append(lstd) brm = td.RecursiveBRMDS(phi=phi) brm.name = "BRMDS" brm.color = "b" brm.ls = "--" methods.append(brm) brm = td.RecursiveBRM(phi=phi) brm.name = "BRM" brm.color = "b" methods.append(brm)
lam = 0.0 eps = 100000 lstd = td.RecursiveLSTDLambda(lam=lam, eps=eps, phi=phi, gamma=gamma) lstd.name = r"LSTD({}) $\epsilon$={}".format(lam, eps) lstd.color = "g" lstd.ls = "-." methods.append(lstd) # alpha = .3 beta = 100. mins = 0 lam = .0 lstd = td.FPKF(lam=lam, alpha=alpha, beta=beta, mins=mins, eps=1, phi=phi, gamma=gamma) lstd.name = r"FPKF({}) $\alpha$={} $\beta={}$".format(lam, alpha, beta) lstd.color = "g" lstd.ls = "-." methods.append(lstd) alpha = .5 rg = td.ResidualGradientDS(alpha=alpha, phi=phi, gamma=gamma) rg.name = r"RG DS $\alpha$={}".format(alpha) rg.color = "brown" rg.ls = "--" methods.append(rg)
lstd = td.RecursiveLSPELambda(lam=lam, alpha=alpha, phi=phi, gamma=gamma) lstd.name = r"LSPE({}) $\alpha$={}".format(lam, alpha) lstd.color = "g" methods.append(lstd) lam = 0. eps = 0.01 lstd = td.RecursiveLSTDLambda(lam=lam, eps=eps, phi=phi, gamma=gamma) lstd.name = r"LSTD({}) $\epsilon$={}".format(lam, eps) lstd.color = "g" lstd.ls = "-." methods.append(lstd) # alpha = 0.0005 lam = .2 lstd = td.FPKF(lam=lam, alpha=alpha, phi=phi, gamma=gamma) lstd.name = r"FPKF({}) $\alpha$={}".format(lam, alpha) lstd.color = "g" lstd.ls = "-." methods.append(lstd) alpha = .0005 rg = td.ResidualGradientDS(alpha=alpha, phi=phi, gamma=gamma) rg.name = r"RG DS $\alpha$={}".format(alpha) rg.color = "brown" rg.ls = "--" methods.append(rg) alpha = .003 rg = td.ResidualGradient(alpha=alpha, phi=phi, gamma=gamma) rg.name = r"RG $\alpha$={}".format(alpha)