def test_hermitian_tuple_scorer():
emb = [[1., 1, 0, 3], [0, 1, 0, 1], [-1, 1, 1, 5]]
tuples_var = tf.Variable([[0, 1], [1, 0], [0, 2], [2, 0], [1, 2], [2, 1]])
(g, params) = hermitian_tuple_scorer(tuples_var, emb0=emb, symmetry_coef=(1.0, 0.0))
print(tf_eval(g))
(g, params) = hermitian_tuple_scorer(tuples_var, emb0=emb, symmetry_coef=(0.0, 1.0))
print(tf_eval(g))
(g, params) = hermitian_tuple_scorer(tuples_var, emb0=emb, symmetry_coef=(0.5, 0.5))
print(tf_eval(g))
(g, params) = hermitian_tuple_scorer(tuples_var, emb0=emb, symmetry_coef=(0.9, 0.1))
print(tf_eval(g)) # close to symmetric
(g, params) = hermitian_tuple_scorer(tuples_var, emb0=emb, symmetry_coef=(0.1, 0.9))
print(tf_eval(g)) # close to anti-symmetric
def test_loss_func_softmax():
"""Test the softmax computation
Returns:
Nothing
>>> test_loss_func_softmax()
[ 2.00384593 2.19245744 0.31507221 1.76036525]
[ 0.20384566 0.59245753 1.31507206 0.26036522]
"""
pred = tf.Variable([[1.1, 2.1, 3.9], [-1.0, -0.5, -2.1], [1.1, 2.1, -3.9], [-1.0, 0.5, -2.1]])
gold = tf.Variable([1, 2, 1, 0])
print(tf_eval(loss_func_softmax(pred, gold)))
gold = tf.Variable([2, 1, 0, 1])
print(tf_eval(loss_func_softmax(pred, gold)))
def test_sparse_relational_hermitian_scoring():
emb = tf.Variable([[1., 1, 0, 3], [0, 1, 0, 1], [-1, 1, 1, 5],
[-3, 1, 0, 2], [-1, 2, -1, -5]])
tuples_var = tf.Variable([[0, 3, 1], [1, 3, 0], [0, 3, 2], [2, 4, 0],
[1, 4, 2], [2, 4, 1]])
g = sparse_relational_hermitian_scoring(emb, tuples_var)
print(tf_eval(g)) # close to anti-symmetric
def test_learning_factorization(verbose=False):
n = 9
m = 8
y_mat = toy_factorization_problem(n=n, m=m, rk=3, noise=1)
rank = 2
batch_size = n * m
tuples = mat2tuples(y_mat)
tuple_iterable = data_to_batches(tuples, minibatch_size=batch_size)
# tuple_iterable = positive_and_negative_tuple_sampler(mat2tuples(y_mat), minibatch_size=batch_size)
sampler, (x, y) = feed_dict_sampler(tuple_iterable, types=[np.int64, np.float32])
loss_op = tf.reduce_mean(loss_func(multilinear_tuple_scorer(x, rank=rank, n_emb=n+m)[0], y, 'quadratic'))
initial_losses = [tf_eval(loss_op, f) for _, f in sampler]
if verbose:
print(initial_losses)
# hooks = [lambda s, e, it, l: it and ((it % 100) == 0) and print("%d) loss=%f" % (it, l))]
hooks = [lambda it, b, l: it and ((it % 1) == 0) and print("{0}) train loss={1}".format(it, l[0]))]
emb, = learn(loss_op, sampler, tf.train.AdamOptimizer(learning_rate=0.1), hooks, max_epochs=200)
# emb, = learn(l, sampler, tf.train.GradientDescentOptimizer(learning_rate=0.5), hooks, max_epochs=500)
# emb, = learn(l, sampler, tf.train.AdagradOptimizer(0.01, initial_accumulator_value=0.01), hooks, max_epochs=500)
mat0 = svd_factorize_matrix(y_mat, rank=2) # exact svd solution
mat = emb[:n, :].dot(emb[n:, :].T)
if verbose:
print(np.linalg.norm(mat - mat0))
assert(np.linalg.norm(mat - mat0) < 1e-3)
def test_hermitian_tuple_scorer():
emb = [[1., 1, 0, 3], [0, 1, 0, 1], [-1, 1, 1, 5]]
tuples_var = tf.Variable([[0, 1], [1, 0], [0, 2], [2, 0], [1, 2], [2, 1]])
(g, params) = hermitian_tuple_scorer(tuples_var,
emb0=emb,
symmetry_coef=(1.0, 0.0))
print(tf_eval(g))
(g, params) = hermitian_tuple_scorer(tuples_var,
emb0=emb,
symmetry_coef=(0.0, 1.0))
print(tf_eval(g))
(g, params) = hermitian_tuple_scorer(tuples_var,
emb0=emb,
symmetry_coef=(0.5, 0.5))
print(tf_eval(g))
(g, params) = hermitian_tuple_scorer(tuples_var,
emb0=emb,
symmetry_coef=(0.9, 0.1))
print(tf_eval(g)) # close to symmetric
(g, params) = hermitian_tuple_scorer(tuples_var,
emb0=emb,
symmetry_coef=(0.1, 0.9))
print(tf_eval(g)) # close to anti-symmetric
def test_learning_factorization(verbose=False):
n = 9
m = 8
y_mat = toy_factorization_problem(n=n, m=m, rk=3, noise=1)
rank = 2
batch_size = n * m
tuples = mat2tuples(y_mat)
tuple_iterable = data_to_batches(tuples, minibatch_size=batch_size)
# tuple_iterable = positive_and_negative_tuple_sampler(mat2tuples(y_mat), minibatch_size=batch_size)
sampler, (x, y) = feed_dict_sampler(tuple_iterable,
types=[np.int64, np.float32])
loss_op = tf.reduce_mean(
loss_func(
multilinear_tuple_scorer(x, rank=rank, n_emb=n + m)[0], y,
'quadratic'))
initial_losses = [tf_eval(loss_op, f) for _, f in sampler]
if verbose:
print(initial_losses)
# hooks = [lambda s, e, it, l: it and ((it % 100) == 0) and print("%d) loss=%f" % (it, l))]
hooks = [
lambda it, b, l: it and (
(it % 1) == 0) and print("{0}) train loss={1}".format(it, l[0]))
]
emb, = learn(loss_op,
sampler,
tf.train.AdamOptimizer(learning_rate=0.1),
hooks,
max_epochs=200)
# emb, = learn(l, sampler, tf.train.GradientDescentOptimizer(learning_rate=0.5), hooks, max_epochs=500)
# emb, = learn(l, sampler, tf.train.AdagradOptimizer(0.01, initial_accumulator_value=0.01), hooks, max_epochs=500)
mat0 = svd_factorize_matrix(y_mat, rank=2) # exact svd solution
mat = emb[:n, :].dot(emb[n:, :].T)
if verbose:
print(np.linalg.norm(mat - mat0))
assert (np.linalg.norm(mat - mat0) < 1e-3)
def test_sparse_relational_hermitian_scoring():
emb = tf.Variable([[1., 1, 0, 3], [0, 1, 0, 1], [-1, 1, 1, 5], [-3, 1, 0, 2], [-1, 2, -1, -5]])
tuples_var = tf.Variable([[0, 3, 1], [1, 3, 0], [0, 3, 2], [2, 4, 0], [1, 4, 2], [2, 4, 1]])
g = sparse_relational_hermitian_scoring(emb, tuples_var)
print(tf_eval(g)) # close to anti-symmetric
