def test_gradients(self):
N = 120
X = np.random.randn(N, 1)
X -= X.mean()
X /= X.std()
Z = X * 0.8 + np.random.randn(N, 1)*0.2
Z += np.abs(Z.min()) + 0.5
Y = np.log(Z)
m = WarpedLMM(Y, X, warping_terms=2)
for i in range(10):
m.randomize()
self.assertTrue(sp.optimize.check_grad(m._f, m._f_prime, m._get_params()) < 1e-4)
def test_model(self):
N = 120
X = np.random.randn(N, 1)
X -= X.mean()
X /= X.std()
Z = X * 0.8 + np.random.randn(N, 1)*0.2
Z += np.abs(Z.min()) + 0.5
Y = np.log(Z)#Z**(1/3.0)
m = WarpedLMM(Y, X, warping_terms=2)
m.randomize()
m.optimize(messages=0)
self.assertTrue(sp.stats.pearsonr(m.Y, Z)[0] >= 0.9)
def test_gradients(self):
N = 120
X = np.random.randn(N, 1)
X -= X.mean()
X /= X.std()
Z = X * 0.8 + np.random.randn(N, 1) * 0.2
Z += np.abs(Z.min()) + 0.5
Y = np.log(Z)
m = WarpedLMM(Y, X, warping_terms=2)
for i in range(10):
m.randomize()
self.assertTrue(
sp.optimize.check_grad(m._f, m._f_prime, m._get_params()) <
1e-4)
def test_model(self):
N = 120
X = np.random.randn(N, 1)
X -= X.mean()
X /= X.std()
Z = X * 0.8 + np.random.randn(N, 1) * 0.2
Z += np.abs(Z.min()) + 0.5
Y = np.log(Z) #Z**(1/3.0)
m = WarpedLMM(Y, X, warping_terms=2)
m.randomize()
m.optimize(messages=0)
self.assertTrue(sp.stats.pearsonr(m.Y, Z)[0] >= 0.9)
