def testConeModelPredictPositiveEpsilon(self):
matrix = np.array([[1,0],[0,1]])
model = ConeModel(matrix, 0.5)
self.assertEqual(model.predict([[0,0],
[0,1],
[1,0],
[1,1]]),
[1,1,1,1])
class ArtificialData:
def __init__(self, data_dims, cone_dims, size=3000):
self.data_dims = data_dims
self.cone_dims = cone_dims
self.size = size
def generate(self, noise=0.0, epsilon=0.0):
rand_array = random.random_sample(self.data_dims*self.cone_dims)*2 - 1
self.cone = rand_array.reshape( (self.cone_dims, self.data_dims) )
self.data = []
self.target = []
cone_inv = pinv(self.cone)
self.model = ConeModel(self.cone, epsilon)
for i in xrange(self.size):
# Generate positive data half the time
if random.random_sample() > 0.5:
v = random.random_sample(self.cone_dims)
v = np.array(np.dot(cone_inv, v))
else:
v = random.random_sample(self.data_dims)*2 - 1.0
if random.random_sample() < noise:
self.data.append(-v)
else:
self.data.append(v)
self.target.append(self.model.predict([v])[0])
self.data = np.array(self.data)
self.target = np.array(self.target)
