bug fix in gp2Scale

fvgp/gp2Scale.py

@@ -97,9 +97,9 @@ def __init__(
                     abs(np.mean(self.y_data) / 100.0)
             print("CAUTION: you have not provided data variances in fvGP,")
             print("they will be set to 1 percent of the data values!")
-        elif variances.dim() == 2:
+        elif np.ndim(variances) == 2:
             self.variances = variances[:,0]
-        elif variances.dim() == 1:
+        elif np.ndim(variances) == 1:
             self.variances = np.array(variances)#, requires_grad = True)
         else:
             raise Exception("Variances are not given in an allowed format. Give variances as 1d numpy array")

fvgp/mcmc.py

@@ -22,19 +22,20 @@ def mcmc(func,bounds, x0 = None, distr = None, max_iter = 1000, ):
     f = []
     if x0 is None: x.append(np.random.uniform(low = bounds[:,0],high = bounds[:,1],size = len(bounds)))
     else: x.append(x0)
-    if distr is None: l = np.diag((np.abs(np.subtract(bounds[:,0],bounds[:,1])))/20.0)**2
+    if distr is None: l = np.diag((np.abs(np.subtract(bounds[:,0],bounds[:,1])))/1000.0)**2
     counter = 0
     current_func = func(x0)
     f.append(current_func)
     run = True
     while run:
         x_proposal = np.random.multivariate_normal(x[-1], l)
-        #x_proposal = project_onto_bounds(x_proposal,bounds)
+        x_proposal = project_onto_bounds(x_proposal,bounds)
         ##check for constraints?
         proposal_func = func(x_proposal) ####call function
         acceptance_prob = proposal_func - current_func ##these are already logs
         u = np.log(np.random.rand())
-        print("current f: ",current_func, " proposal f: ",proposal_func, "acceptance probability: ",acceptance_prob," u ",u, flush = True)
+        print("iteration: ", counter,"current f: ",current_func, " prop f: ",proposal_func,flush = True)
+        print("acceptance prob: ",acceptance_prob,"u: ", u,flush = True)
         if u < acceptance_prob:
             x.append(x_proposal)
             f.append(proposal_func)
@@ -45,6 +46,7 @@ def mcmc(func,bounds, x0 = None, distr = None, max_iter = 1000, ):
             f.append(current_func)
         print(f[-1], flush = True)
         logger.debug("mcmc f(x):{}",f[-1])
+        print("")
         counter += 1
         if counter >= max_iter: run = False
         #if len(x)>201 and np.linalg.norm(np.mean(x[-100:],axis = 0)-np.mean(x[-200:-100],axis = 0)) < 0.01 * np.linalg.norm(np.mean(x[-100:],axis = 0)): run = False

fvgp/sparse_matrix.py

@@ -126,7 +126,6 @@ def random_logdet(self, A,eps = 10.0,delta = 0.1,m = 100):
         p = int(20.0 * np.log(2./delta)/eps**2)+1
         gamma = np.zeros((p,m))
         for i in range(p):
-            print(i," of ",p)
             g = np.random.normal(0, 1., size = N)
             v = C @ g
             gamma[i,1] = g.T @ v