}
else:
kwargs = {
'D': D,
'M': M,
'learning_rate': args.alpha,
'F': [(1, 1)],
'c0': c0,
'C': [(c0, )]
}
graph_f = cnn.build_graph
else:
kwargs = {'D': D, 'M': M, 'learning_rate': args.alpha}
graph_f = ann.build_graph
pol = EpsilonGreedyPolicy(eps=1.0, decay_f=decay_f)
pol.n = args.eps_start
if args.mode == 'leaf':
sv = tdleaf.TDLeafSupervisor(pol,
mv_limit=args.move_count,
depth=args.depth,
y=args.gamma,
l=args.lambd)
else:
sv = tdstem.TDStemSupervisor(pol,
mv_limit=args.move_count,
depth=args.depth,
y=args.gamma,
l=args.lambd)
sv.run(args.I,
args.N,
graph_f,
