def test_ess_sampling(self, dtype, device, lap, M):
try:
k = 2
g = rand_udg(N, dtype=dtype, device=device)
L = laplace(g, lap_type=lap)
S = ess(L, M, k)
print(S)
except ArpackNoConvergence:
print("No convergence error")
def test_computing_sets(dtype):
g = rand_udg(N, 0.1, dtype=dtype)
sets, lengths = computing_sets(g, T=8e-5, p_hops=3)
print(sets)
print(lengths)
sets, lengths = computing_sets(g.set_value_(None), T=8e-5, p_hops=3)
print(sets)
print(lengths)
def test_recon(self, dtype, device, lap, M):
g = rand_udg(N, dtype=dtype, device=device)
L = laplace(g, lap_type=lap)
S = ess(L, M)
num_sig = 10
f = th.rand(N, num_sig, dtype=dtype, device=device)
fs = f[S, :]
f_hat = recon_ess(fs, S, g.U(lap), bd=int(3 * N / 4))
snr_and_mse(f_hat, f)
def test_recon_bsgda():
mu = 0.01
g = rand_udg(N, 0.1, dtype=torch.double)
sampled_nodes, thresh = bsgda(g, K * 8, mu=mu, epsilon=1e-8)
print("sampled nodes: ", sampled_nodes)
print("thresh(T): ", thresh)
f = torch.randn(N, 100, dtype=torch.double)
f_hat = recon_bsgda(f[sampled_nodes], sampled_nodes, g.L("comb"), mu)
m, s = snr_and_mse(f_hat, f)
def test_consistency_greedy_sampling(dtype):
g = rand_udg(N, 0.3, dtype=dtype).set_value_(None)
selected_pebbles, vf = greedy_sampling(g, K, T, p_hops=max_hops)
sets, lengths = computing_sets(g, T, p_hops=max_hops)
s1, vf1 = solving_set_covering(sets, lengths, K)
s2, vf2 = greedy_gda_sampling(g, K, T, p_hops=max_hops)
print(selected_pebbles)
print(s1)
print(s2)
print(vf)
print(vf1)
print(vf2)
assert s1 == selected_pebbles
assert vf1 == vf
assert s1 == s2
assert vf1 == vf2
def test_bsgda(dtype):
g = rand_udg(N, 0.1, dtype=dtype).fill_value_(1.0)
sampled_nodes, thresh = bsgda(g, K * 2)
print("sampled nodes: ", sampled_nodes)
print("thresh(T): ", thresh)