for link in G:
times += G[link]
if test:
times_test = []
for link in G_test:
times_test += G_test[link]
times = np.sort(times)
if test:
times_test = np.sort(times_test)
## Build the MEG model
m = meg.meg_model(G,
verbose=False,
tau_zero=tau_zero,
full_links=full_links,
discrete=True)
m.specification(main_effects=main_effects,
interactions=interactions,
poisson_me=poisson_main_effects,
poisson_int=poisson_interactions,
hawkes_me=hawkes_main_effects,
hawkes_int=hawkes_interactions,
D=D,
verbose=False)
np.random.seed(117711)
m.prior_initialisation()
## Initialise all to the same values
init_i = np.array([Counter(m.ni)[x] / m.T / m.n + 1e-9 for x in range(m.n)])
## Empty matrix
A = {}
for i in range(n):
for j in range(n):
A[i, j] = [1]
## Simulate
np.random.seed(S)
G = {}
for i in range(nsim):
## MEG model setup for simulation
m = meg.meg_model(A,
tau_zero=True,
full_links=True,
verbose=False,
discrete=False,
force_square=True,
evaluate_directed=False)
m.specification(main_effects=False,
interactions=True,
poisson_me=False,
poisson_int=False,
hawkes_me=True,
hawkes_int=True,
D=1,
verbose=False)
m.prior_initialisation()
## True values of the parameters
m.gamma = np.array([0.1, 0.5])
m.gamma_prime = np.array([0.1, 0.3])
## Remove repeated observations
times = np.sort(np.unique([float(x) for x in line[2:]]))
times = times[times > tmin] - tmin
times_test = times[np.logical_and(times >= 86400 * 14,
times < 86400 * 21)]
if comp in huxley_map and host in server_map and len(
times_test) > 0:
if (huxley_map[comp], server_map[host]) in G_test:
G_test[huxley_map[comp],
server_map[host]] += list(times_test)
else:
G_test[huxley_map[comp],
server_map[host]] = list(times_test)
## Build the MEG model
m = meg.meg_model(G, verbose=False)
m.specification(main_effects=main_effects,
interactions=interactions,
poisson_me=poisson_main_effects,
poisson_int=poisson_interactions,
hawkes_me=hawkes_main_effects,
hawkes_int=hawkes_interactions,
D=D,
verbose=False)
np.random.seed(117711)
## Prior initialisation
m.prior_initialisation()
## Initialise
init_i = np.array([Counter(m.ni)[x] / m.T / m.n1 + 1e-9 for x in range(m.n1)])
init_j = np.array(
hawkes_main_effects = True if args.hawkes_main_effects else False
hawkes_interactions = True if args.hawkes_interactions else False
D = args.d
eta = args.eta
N_nodes = args.n
## Parse arguments
G = {}
j = 0
A = np.load(input_folder + '/graphs.npy', allow_pickle='TRUE').item()
for index in A:
G[j] = A[index]
j += 1
## Model
m = meg.meg_model(G[0], tau_zero=True, verbose=False, discrete=False, force_square=True)
## Extract n
m.n = N_nodes
## Repeat initialisations 5 times
ks_scores = []; ks_pvals = []
if main_effects:
alpha = np.zeros((j,N_nodes)); beta = np.zeros((j,N_nodes))
if not poisson_main_effects:
mu = np.zeros((j,N_nodes)); phi = np.zeros((j,N_nodes))
mu_prime = np.zeros((j,N_nodes)); phi_prime = np.zeros((j,N_nodes))
if interactions:
if D == 1:
gamma = np.zeros((j,N_nodes)); gamma_prime = np.zeros((j,N_nodes))
nu = np.zeros((j,N_nodes)); nu_prime = np.zeros((j,N_nodes))