dimensionality bug fix

Author: erdem-biyik

GP.py

@@ -13,10 +13,9 @@ def __init__(self, initialPoint=0, theta=0.1, noise_level=0.1):
         self.theta = theta #hyperparameter
         self.W = np.zeros((2,2)) #hessian at the queries
         self.noise = noise_level
-        self.initialPoint = np.array([initialPoint])  #initial point that is set to have a 0 value
+        self.initialPoint = np.array(initialPoint)  #initial point that is set to have a 0 value
         self.dim = len(self.initialPoint) #number of features
 
-
     def updateParameters(self,query,answer):
         self.listQueries.append([query[0],query[1],answer])
         self.K = self.covK()

test.py

@@ -5,7 +5,7 @@
 def findBestQuery(gp):
     def negative_info_gain(x):
         return -1*gp.objectiveEntropy(x)
-    x0 = np.array(gp.initialPoint*2) + np.random.rand(gp.dim*2)
+    x0 = np.array(list(gp.initialPoint)*2) + np.random.rand(gp.dim*2)
     # Let's now find the optimal query within the bounds (-2,2) for each dimension
     opt_res = opt.fmin_l_bfgs_b(negative_info_gain, x0=x0, bounds=[(-2,2)]*gp.dim*2, approx_grad=True, iprint=-1)
     return opt_res[0], -opt_res[1]
@@ -21,7 +21,7 @@ def negative_info_gain(x):
 print('posterior mean for the feature set [3,1] = ' + str(gp.mean1pt([3,1])))
 print('posterior covariance between the features [4,0] and [-1,1] = ' + str(gp.postcov([4,0],[-1,1])))
 print('posterior variance of the feature set [-2,1] = ' + str(gp.cov1pt([-2,1])))
-print('expected information gain from the query [0,0] vs [2,2] = ' + str(gp.objectiveEntropy([[0,0],[2,2]])))
+print('expected information gain from the query [0,0] vs [2,2] = ' + str(gp.objectiveEntropy([0,0,2,2])))
 
 optimal_query, info_gain = findBestQuery(gp)
 print('optimal next query is ' + str(optimal_query))