def test_normalization():
hom = markov.MarkovChain(3, 1)
hom2 = markov.MarkovChain(3, 2)
hom3 = markov.MarkovChain(3, 3)
hom.fit(MOCK_DATA_RANDOM)
hom2.fit(MOCK_DATA_RANDOM)
hom3.fit(MOCK_DATA_RANDOM)
sums = hom.transition_matrix.sum(axis=1)
sums2 = hom2.transition_matrix.sum(axis=1)
sums3 = hom3.transition_matrix.sum(axis=1)
# some rows might be all zeros...
assert all(i == pytest.approx(1, 0.01) for i in sums if i)
assert all(i == pytest.approx(1, 0.01) for i in sums2 if i)
assert all(i == pytest.approx(1, 0.01) for i in sums3 if i)
def test_possible_states_order_2_small():
hom = markov.MarkovChain(3, 2)
hom._update_transition_matrix(MOCK_SEQUENCE_SMALL)
assert hom.possible_states == EXPECTED_POSSIBLE_STATES_ORDER_2_SMALL
df = hom.transition_df()
assert set(df.columns.values) == set(hom.possible_states.keys())
assert set(df.index.values) == set(hom.possible_states.keys())
def test_evolve_states_high_order():
hom = markov.MarkovChain(3, 2)
hom.fit(MOCK_SEQUENCE_SMALL)
initial_state = np.zeros(3)
initial_state[0] = 1
vec = hom.evolve_states(initial_state, num_steps=1, threshold=0)
def test_next_state():
hom = markov.MarkovChain(3, 1)
hom.fit(MOCK_DATA_RANDOM)
# transition: 0.33583959899749372, 0.34335839598997492, 0.32080200501253131
# taken from the initial transition matrix after training
initial_state = np.array(
[0.33583959899749372, 0.34335839598997492, 0.32080200501253131])
next_state = hom.predict_state(initial_state, num_steps=2)
assert next_state[0, 1] == pytest.approx(0.341, 0.01)
def test_next_state_high_order():
hom = markov.MarkovChain(3, 2)
hom.fit(MOCK_DATA_RANDOM)
# take a row from the initial transition matrix after training
# as initial state.
initial_state = hom.transition_matrix.todense()[1, :]
next_state = hom.predict_state(initial_state, num_steps=2)
assert next_state[0, 1] == pytest.approx(0.111, 0.01)
def test_build_pos():
hom = markov.MarkovChain(3, 1)
hom.fit(MOCK_SEQUENCE_SMALL)
initial_state = np.zeros(3)
initial_state[0] = 1
vec = hom.evolve_states(initial_state, num_steps=1, threshold=0)
pos = hom.build_pos(vec)
assert {0: (0, 0), 1: (1, 0), 2: (1, -1)} == pos
def test_build_graph():
hom = markov.MarkovChain(3, 1)
hom.fit(MOCK_SEQUENCE_SMALL)
initial_state = np.zeros(3)
initial_state[0] = 1
vec = hom.evolve_states(initial_state, num_steps=2, threshold=0)
graph = hom.generate_graph(vec)
assert graph
assert len(graph.nodes()) == 8
assert len(graph.edges()) == 7
def test_evolve_states_more_steps():
hom = markov.MarkovChain(3, 1)
hom.fit(MOCK_SEQUENCE_SMALL)
initial_state = np.zeros(3)
initial_state[0] = 1
vec = hom.evolve_states(initial_state, num_steps=2, threshold=0)
assert vec[2][0]["actual"] == pytest.approx(0.64, 0.01)
assert vec[2][1]["actual"] == pytest.approx(0.16, 0.01)
assert vec[2][2]["actual"] == pytest.approx(0.015, 0.05)
assert vec[2][3]["actual"] == pytest.approx(0.09, 0.05)
assert vec[2][4]["actual"] == pytest.approx(0.09, 0.05)
vec1 = hom.evolve_states(initial_state, num_steps=2, threshold=0.1)
assert len(vec1[2]) == 2 # only those above threshold
def test_evolve_states():
hom = markov.MarkovChain(3, 1)
hom.fit(MOCK_SEQUENCE_SMALL)
initial_state = np.zeros(3)
initial_state[0] = 1
vec = hom.evolve_states(initial_state, num_steps=1)
assert vec[0][0]["state_id"] == 0
assert vec[0][0]["state"] == 0
assert vec[0][0]["weight"] == 1
assert vec[0][0]["actual"] == 1
assert len(vec[1]) == 2
assert vec[1][0]["actual"] == hom.transition_matrix[0, 0]
assert vec[1][1]["actual"] == hom.transition_matrix[0, 2]
def test_possible_states_order_2():
hom = markov.MarkovChain(6, 2)
hom._update_transition_matrix(MOCK_SEQUENCE)
assert hom.possible_states == EXPECTED_POSSIBLE_STATES_ORDER_2