testing #19 for gpr aep -- not working

thangbui · Jun 12, 2017 · 68b6886 · 68b6886
1 parent cf7910c
commit 68b6886
Show file tree

Hide file tree

Showing 3 changed files with 11 additions and 12 deletions.
diff --git a/examples/gpr_aep_examples.py b/examples/gpr_aep_examples.py
@@ -9,7 +9,7 @@
 from .datautils import step, spiral
 from .context import aep
 
-def run_regression_1D():
+def run_regression_1D(nat_param=True):
     np.random.seed(42)
 
     print "create dataset ..."
@@ -37,10 +37,12 @@ def plot(m):
     # inference
     print "create model and optimize ..."
     M = 20
-    model = aep.SGPR(X, Y, M, lik='Gaussian')
-    model.optimise(method='L-BFGS-B', alpha=0.1, maxiter=50000)
-    plot(model)
-    plt.show()
+    model = aep.SGPR(X, Y, M, lik='Gaussian', nat_param=nat_param)
+    model.update_hypers(model.init_hypers(Y))
+    print model.objective_function(model.get_hypers(), N, 0.1)
+    model.optimise(method='L-BFGS-B', alpha=0.1, maxiter=1)
+    # plot(model)
+    # plt.show()
 
 
 def run_banana():
@@ -307,11 +309,12 @@ def run_boston():
 
 
 if __name__ == '__main__':
-    # run_regression_1D()
+    run_regression_1D(True)
+    run_regression_1D(False)
     # run_banana()
     # run_step_1D()
     # run_spiral()
     # run_boston()
 
     # run_regression_1D_stoc()
-    run_banana_stoc()
+    # run_banana_stoc()
diff --git a/geepee/base_models.py b/geepee/base_models.py
@@ -505,9 +505,6 @@ def compute_posterior_grad_u(self, dmu, dSu):
 
         return deta1_R, deta2, dKuuinv
 
-
-
-
     def init_hypers(self, x_train=None, key_suffix=''):
         """Summary
         
@@ -575,7 +572,6 @@ def init_hypers(self, x_train=None, key_suffix=''):
             triu_ind = np.triu_indices(M)
             diag_ind = np.diag_indices(M)
             R[diag_ind] = np.log(R[diag_ind])
-            np.log(R[diag_ind])
             eta1_d = R[triu_ind].reshape((M * (M + 1) / 2,))
             eta2_d = theta2.reshape((M,))
             eta1_R[d, :] = eta1_d

diff --git a/tests/test_grads_aep.py b/tests/test_grads_aep.py
@@ -5135,7 +5135,7 @@ def plot_gpssm_linear_aep_gaussian_stochastic():
     # plot_gplvm_aep_probit_stochastic()
     # plot_gplvm_aep_gaussian_stochastic()
 
-    test_gpr_aep_gaussian(True)
+    # test_gpr_aep_gaussian(True)
     test_gpr_aep_gaussian(False)
     # test_gpr_aep_probit()
     # test_gpr_aep_gaussian_scipy()