def test_setWeights(self):
rbm = DeepRBM([2, 3, 1])
with self.assertRaises(ValueError):
rbm.setWeights([
np.array([[0, 0], [0, 0], [0, 0]]),
np.array([[0], [0], [0]])
])
rbm.setWeights(
[np.array([[0, 0], [0, 0], [0, 0]]),
np.matrix([0, 0, 0])])
def test_constructor(self):
drbm = DeepRBM([5, 2, 3])
self.assertEqual(len(drbm.weights), 2)
self.assertEqual(drbm.weights[0].shape[0], 2)
self.assertEqual(drbm.weights[0].shape[1], 5)
self.assertEqual(drbm.weights[1].shape[0], 3)
self.assertEqual(drbm.weights[1].shape[1], 2)
def test_setWeights(self):
rbm = DeepRBM([2, 3, 1])
#wrong should throw error
with self.assertRaises(ValueError):
rbm.setWeights([np.array([0,0,0],[0,0,0]), np.array([0,0,0])])
rbm.setWeights([np.array([0,0,0],[0,0,0]), np.array([0,0,0]).reshape((1, 3))])
def test_sample(self):
rbm = DeepRBM([2, 3, 1])
rbm.setWeights([np.array([[1,1],[1,1],[1,1]]), np.matrix([1,1,1])])
result = rbm.sample(np.array([0,0]), 0, 2, False)
self.assertAlmostEqual(result[0], 0.8175745, 4)
result = rbm.sample(np.array([[0,0],[1,1],[0,1]]), 0, 2, False)
self.assertAlmostEqual(result[0], 0.8175745, 4)
self.assertAlmostEqual(result[1], 0.933540476818325, 4)
self.assertAlmostEqual(result[2], 0.8996350, 4)
result = rbm.sample(np.array([0]), 2, 0, False)
self.assertAlmostEqual(result[0,0], 0.8175745, 4)
self.assertAlmostEqual(result[0,1], 0.8175745, 4)
def test_setWeights(self):
rbm = DeepRBM([2, 3, 1])
with self.assertRaises(ValueError):
rbm.setWeights([np.array([[0,0],[0,0],[0,0]]), np.array([[0],[0],[0]])])
rbm.setWeights([np.array([[0,0],[0,0],[0,0]]), np.matrix([0,0,0])])
