def test_rmargll_after_optimization(self):
model = LaplaceGLMM_N3K1('logistic')
model.setG(self._G0, self._G1)
model.setX(self._X)
model.sety(self._y)
model.sig02 = 1.5
model.sig12 = 0.5
model.sign2 = 0.8
model.beta = NP.array([2.0, -1.0])
model.optimize()
self.assertAlmostEqual(-6.75677866924,
model._regular_marginal_loglikelihood())
model = EPGLMM_N3K1('erf')
model.setG(self._G0, self._G1)
model.setX(self._X)
model.sety(self._y)
model.sig02 = 1.5
model.sig12 = 0.5
model.sign2 = 0.8
model.beta = NP.array([2.0, -1.0])
model.optimize()
self.assertAlmostEqual(-6.75424440734,
model._regular_marginal_loglikelihood())
model = LaplaceGLMM_N3K1('logistic', 'l2')
model.setG(self._G0, self._G1)
model.setX(self._X)
model.sety(self._y)
model.sig02 = 1.5
model.sig12 = 0.5
model.sign2 = 0.8
model.beta = NP.array([2.0, -1.0])
model.optimize()
self.assertAlmostEqual(-6.7584132443836662,
model._regular_marginal_loglikelihood())
model = EPGLMM_N3K1('erf', 'l2')
model.setG(self._G0, self._G1)
model.setX(self._X)
model.sety(self._y)
model.sig02 = 1.5
model.sig12 = 0.5
model.sign2 = 0.8
model.beta = NP.array([2.0, -1.0])
model.optimize()
self.assertAlmostEqual(-6.7545709287754718,
model._regular_marginal_loglikelihood())
def test_prediction(self):
model = LaplaceGLMM_N3K1('logistic')
model.setG(self._G0, self._G1)
model.setX(self._X)
model.sety(self._y)
model.sig02 = 1.5
model.sig12 = 0.5
model.sign2 = 0.8
model.beta = NP.array([2.0, -1.0])
ps = NP.array([0.30481858,0.22520247,0.78060709,0.44734337,0.58824651,0.052388,\
0.60337951,0.22886631,0.54169641,0.71888192])
p = model.predict(self._X, self._G0, self._G1)
for i in range(ps.shape[0]):
self.assertAlmostEqual(ps[i], p[i], places=4)
model = EPGLMM_N3K1('erf')
model.setG(self._G0, self._G1)
model.setX(self._X)
model.sety(self._y)
model.sig02 = 1.5
model.sig12 = 0.5
model.sign2 = 0.8
model.beta = NP.array([2.0, -1.0])
ps = NP.array([0.35487175, 0.20959901, 0.77285232, 0.48305102, 0.59542164,\
0.02770879, 0.6205743 , 0.14239838, 0.45330375, 0.77123692])
p = model.predict(self._X, self._G0, self._G1)
for i in range(ps.shape[0]):
self.assertAlmostEqual(ps[i], p[i])
def test_setK_rmargll_after_optimization(self):
model = LaplaceGLMM_N3K1('logistic')
model.setK(NP.dot(self._G0, self._G0.T), NP.dot(self._G1, self._G1.T))
model.setX(self._X)
model.sety(self._y)
model.sig02 = 1.5
model.sig12 = 0.5
model.sign2 = 0.8
model.beta = NP.array([2.0, -1.0])
model.optimize()
self.assertAlmostEqual(-6.75677866924,
model._regular_marginal_loglikelihood())
model = EPGLMM_N3K1('erf')
model.setK(NP.dot(self._G0, self._G0.T), NP.dot(self._G1, self._G1.T))
model.setX(self._X)
model.sety(self._y)
model.sig02 = 1.5
model.sig12 = 0.5
model.sign2 = 0.8
model.beta = NP.array([2.0, -1.0])
model.optimize()
self.assertAlmostEqual(-6.75424440734,
model._regular_marginal_loglikelihood())
def test_margll_gradient(self):
def func(x, model):
model.sig02 = x[0]
model.sig12 = x[1]
model.sign2 = x[2]
model.beta = x[3:]
return model.marginal_loglikelihood()
def grad(x, model):
model.sig02 = x[0]
model.sig12 = x[1]
model.sign2 = x[2]
model.beta = x[3:]
return model._mll_gradient()
model = LaplaceGLMM_N3K1('logistic', 'l2')
model.setG(self._G0, self._G1)
model.setX(self._X)
model.sety(self._y)
self.assertAlmostEqual( 0.0, SP.optimize.check_grad(func, grad, NP.array([1.0,1.2,1.5,-3.0,0.4]), model), places=5 )
model = EPGLMM_N3K1('erf', 'l2')
model.setG(self._G0, self._G1)
model.setX(self._X)
model.sety(self._y)
self.assertAlmostEqual( 0.0, SP.optimize.check_grad(func, grad, NP.array([1.0,1.2,1.5,-3.0,0.4]), model), places=5 )
def test_margll(self):
model = LaplaceGLMM_N3K1('logistic', 'l2')
model.setG(self._G0, self._G1)
model.setX(self._X)
model.sety(self._y)
model.sig02 = 1.5
model.sig12 = 0.5
model.sign2 = 0.8
model.beta = NP.array([2.0, -1.0])
self.assertAlmostEqual(-14.635329282675063, model.marginal_loglikelihood())
model = EPGLMM_N3K1('erf', 'l2')
model.setG(self._G0, self._G1)
model.setX(self._X)
model.sety(self._y)
model.sig02 = 1.5
model.sig12 = 0.5
model.sign2 = 0.8
model.beta = NP.array([2.0, -1.0])
self.assertAlmostEqual(-17.413719862714526, model.marginal_loglikelihood())
def test_rmargll(self):
model = LaplaceGLMM_N3K1('logistic')
model.setG(self._G0, self._G1)
model.setX(self._X)
model.sety(self._y)
model.sig02 = 1.5
model.sig12 = 0.5
model.sign2 = 0.8
model.beta = NP.array([2.0, -1.0])
self.assertAlmostEqual(-13.821329282675068, model._regular_marginal_loglikelihood())
model = EPGLMM_N3K1('erf')
model.setG(self._G0, self._G1)
model.setX(self._X)
model.sety(self._y)
model.sig02 = 1.5
model.sig12 = 0.5
model.sign2 = 0.8
model.beta = NP.array([2.0, -1.0])
self.assertAlmostEqual(-16.599719862714529, model._regular_marginal_loglikelihood())