def test_linreg_20_test(self):
X, y = gen_regression_data()
args = Args(data_set="./out/lin.p",
solution="./out/lin-sol.p",
engine="linreg")
mlak.set_preferred_engine(args.engine)
with CapturedStdout() as out:
mlak.test(args)
self.assertAlmostEqual(float(out.getvalue()), 7.117094744250965e-06)
def test_keras_20_test(self):
X, y = gen_regression_data()
args = Args(data_set="./out/keras.p",
solution="./out/keras-sol.p",
engine="keras")
mlak.set_preferred_engine(args.engine)
with CapturedStdout() as out:
mlak.test(args)
self.assertEqual(out.getvalue(), "0.0\n")
def test_struggle_10_train(self):
X, y = gen_logistic_data()
dio.save("./out/struggle.p", {"X": X, "y": y})
args = Args(data_set="./out/struggle.p",
solution="./out/struggle-sol.p",
topology=["10"],
Lambda="0.1",
iterations=1,
engine="struggle")
mlak.set_preferred_engine(args.engine)
with CapturedStdout():
mlak.train(args)
with CapturedStdout() as out:
mlak.test(args)
self.assertEqual(out.getvalue(), "0.5706666666666667\n")
solution = dio.load("./out/struggle-sol.p")
npt.assert_equal(solution.shaper()._classesIdToLabel,
["diamond", "drill", "ripple"])
npt.assert_equal(solution.model()[0], [256, 10, 3])
t = solution.model()[1]
npt.assert_almost_equal([np.sum(
t[0]), np.sum(t[1]), np.sum(t[2])],
[0.14382, 0.0245429, -0.132043])
self.assertAlmostEqual(np.sum(solution.shaper().mu()),
205.85431033692913)
self.assertAlmostEqual(np.sum(solution.shaper().sigma()),
296.5719143203778)
fix_random()
args.iterations = 9
with CapturedStdout():
mlak.train(args)
with CapturedStdout() as out:
mlak.test(args)
self.assertEqual(out.getvalue(), "0.0\n")
solution = dio.load("./out/struggle-sol.p")
npt.assert_equal(solution.shaper()._classesIdToLabel,
["diamond", "drill", "ripple"])
npt.assert_equal(solution.model()[0], [256, 10, 3])
t = solution.model()[1]
npt.assert_almost_equal([np.sum(
t[0]), np.sum(t[1]), np.sum(t[2])],
[0.3095417, -0.0072933, -0.1675103])
self.assertAlmostEqual(np.sum(solution.shaper().mu()),
205.85431033692913)
self.assertAlmostEqual(np.sum(solution.shaper().sigma()),
296.5719143203778)
def test_logreg_10_train(self):
X, y = gen_logistic_data()
dio.save("./out/log.p", {"X": X, "y": y})
args = Args(data_set="./out/log.p",
solution="./out/log-sol.p",
topology=[],
Lambda="0.1",
iterations=1,
engine="logreg")
mlak.set_preferred_engine(args.engine)
with CapturedStdout():
mlak.train(args)
with CapturedStdout() as out:
mlak.test(args)
self.assertEqual(out.getvalue(), "0.2666666666666667\n")
solution = dio.load("./out/log-sol.p")
npt.assert_equal(solution.shaper()._classesIdToLabel,
["diamond", "drill", "ripple"])
self.assertEqual(len(solution.model()), 3)
npt.assert_almost_equal(np.sum(solution.model(), axis=1),
[15.9547132, -0.9575446, -4.9745893])
self.assertAlmostEqual(np.sum(solution.shaper().mu()),
205.85431033692913)
self.assertAlmostEqual(np.sum(solution.shaper().sigma()),
296.5719143203778)
fix_random()
args.iterations = 9
with CapturedStdout():
mlak.train(args)
with CapturedStdout() as out:
mlak.test(args)
self.assertEqual(out.getvalue(), "0.0\n")
solution = dio.load("./out/log-sol.p")
npt.assert_equal(solution.shaper()._classesIdToLabel,
["diamond", "drill", "ripple"])
self.assertEqual(len(solution.model()), 3)
npt.assert_almost_equal(np.sum(solution.model(), axis=1),
[20.5073757, -13.1428356, -10.8690069])
self.assertAlmostEqual(np.sum(solution.shaper().mu()),
205.85431033692913)
self.assertAlmostEqual(np.sum(solution.shaper().sigma()),
296.5719143203778)