add_network_stats(s_results)
s_results.new('Subnetwork kappa', 'm', lambda d, f: d.base_model.kappa)
def f_c(c):
return ((lambda d, f: d.base_model.beta[c]),
(lambda d, f: f.base_model.beta[c]))
for c in covariates:
# Need to do this hackily to avoid for-loop/lambda-binding weirdness.
f_true, f_estimated = f_c(c)
s_results.new('True beta_{%s}' % c, 'm', f_true)
s_results.new('Estimated beta_{%s}' % c, 'm', f_estimated)
s_results.new('Class mismatch', 'n',
lambda n: minimum_disagreement(n.node_covariates['z_true'][:], \
n.node_covariates['z'][:]))
def rel_mse_p_ij(n, d, f):
P = d.edge_probabilities(n)
return rel_mse(f.edge_probabilities(n), f.baseline(n), P)
s_results.new('Rel. MSE(P)', 'nm', rel_mse_p_ij)
def rel_mse_logit_p_ij(n, d, f):
logit_P = logit(d.edge_probabilities(n))
logit_Q = f.baseline_logit(n)
return rel_mse(logit(f.edge_probabilities(n)), logit_Q, logit_P)
# Calculate NLL at initialized block assignments
fit_model.fit_sem(net, cycles=1, sweeps=0, use_best=False, store_all=True)
baseline_nll = fit_model.sem_trace[0][0]
nll_trace = []
z_trace = np.empty((steps, N))
disagreement_trace = []
theta_trace = []
for step in range(steps):
print step
fit_model.fit_sem(net, 1, 2, store_all=True)
#fit_model.fit_kl(net, 1)
nll_trace.append(fit_model.nll(net))
z_trace[step, :] = net.node_covariates['z'][:]
disagreement = minimum_disagreement(net.node_covariates['value'][:],
net.node_covariates['z'][:])
disagreement_trace.append(disagreement)
theta_trace.append(fit_model.base_model.beta['x'])
# Eliminate symmetry of 'z'
for step in range(steps):
if np.mean(z_trace[step, :]) < 0.5:
z_trace[step, :] = 1 - z_trace[step, :]
z_trace += np.random.normal(0, 0.01, (steps, N))
nll_trace = np.array(nll_trace)
nll_trace -= baseline_nll
disagreement_trace = np.array(disagreement_trace)
plt.figure()
plt.plot(np.arange(steps), theta_trace)
net.new_node_covariate_int('z')
# Set up recording of results from experiment
s_results = Results(params['sub_sizes'], params['num_reps'], 'Stationary fit')
add_network_stats(s_results)
s_results.new('Subnetwork kappa', 'm', lambda d, f: d.base_model.kappa)
def f_c(c):
return ((lambda d, f: d.base_model.beta[c]),
(lambda d, f: f.base_model.beta[c]))
for c in covariates:
# Need to do this hackily to avoid for-loop/lambda-binding weirdness.
f_true, f_estimated = f_c(c)
s_results.new('True beta_{%s}' % c, 'm', f_true)
s_results.new('Estimated beta_{%s}' % c, 'm', f_estimated)
s_results.new('Class mismatch', 'n',
lambda n: minimum_disagreement(n.node_covariates['z_true'][:], \
n.node_covariates['z'][:]))
def rel_mse_p_ij(n, d, f):
P = d.edge_probabilities(n)
return rel_mse(f.edge_probabilities(n), f.baseline(n), P)
s_results.new('Rel. MSE(P)', 'nm', rel_mse_p_ij)
def rel_mse_logit_p_ij(n, d, f):
logit_P = logit(d.edge_probabilities(n))
logit_Q = f.baseline_logit(n)
return rel_mse(logit(f.edge_probabilities(n)), logit_Q, logit_P)
s_results.new('Rel. MSE(logit_P)', 'nm', rel_mse_logit_p_ij)
all_results = { 's': s_results }
if params['fit_nonstationary']:
n_results = s_results.copy()
n_results.title = 'Nonstationary fit'
all_results['n'] = n_results
def class_mismatch(n):
truth = n.node_covariates['truth'][:]
estimated = n.node_covariates['z'][:]
return minimum_disagreement(truth, estimated, normalized = False)
fit_model.fit_sem(net, cycles = 1, sweeps = 0,
use_best = False, store_all = True)
baseline_nll = fit_model.sem_trace[0][0]
nll_trace = []
z_trace = np.empty((steps,N))
disagreement_trace = []
theta_trace = []
for step in range(steps):
print step
fit_model.fit_sem(net, 1, 2, store_all = True)
#fit_model.fit_kl(net, 1)
nll_trace.append(fit_model.nll(net))
z_trace[step,:] = net.node_covariates['z'][:]
disagreement = minimum_disagreement(net.node_covariates['value'][:],
net.node_covariates['z'][:])
disagreement_trace.append(disagreement)
theta_trace.append(fit_model.base_model.beta['x'])
# Eliminate symmetry of 'z'
for step in range(steps):
if np.mean(z_trace[step,:]) < 0.5:
z_trace[step,:] = 1 - z_trace[step,:]
z_trace += np.random.normal(0, 0.01, (steps, N))
nll_trace = np.array(nll_trace)
nll_trace -= baseline_nll
disagreement_trace = np.array(disagreement_trace)
plt.figure()
plt.plot(np.arange(steps), theta_trace)
