def test_weights_have_correct_sequence():
weights = generate_weights(NeuronCount(input=2, hidden=4))
quarters = quarter(weights)
assert np.all(quarters.first > 0)
assert np.all(quarters.second < 0)
assert np.all(quarters.third < 0)
assert np.all(quarters.fourth > 0)
def test_weights_are_in_range(input_neuron_count, hidden_neuron_count):
weights = generate_weights(
NeuronCount(input=input_neuron_count, hidden=hidden_neuron_count))
min_weight = min(POSSIBLE_ABSOLUTE_WEIGHTS)
max_weight = max(POSSIBLE_ABSOLUTE_WEIGHTS)
for weight in weights:
assert min_weight <= weight <= max_weight or -max_weight <= weight <= -min_weight
def test_weights_have_correct_sequence_with_multiple_input_pairs():
weights = generate_weights(NeuronCount(input=4, hidden=32))
half_point = len(weights) // 2
for weight_half in [weights[:half_point], weights[half_point:]]:
quarters = quarter(weight_half)
assert np.all(quarters.first > 0)
assert np.all(quarters.second < 0)
assert np.all(quarters.third < 0)
assert np.all(quarters.fourth > 0)
def _read_neuron_counts(configuration):
input_count = _measurement_mode_to_input_count(
configuration.measurement_mode)
return NeuronCount(input_count, configuration.hidden_neuron_count)
def test_weight_generation_fails_on_odd_number_of_hidden_neurons():
with pytest.raises(ValueError):
generate_weights(NeuronCount(input=2, hidden=3))
def test_weights_sum_to_zero(input_neuron_count, hidden_neuron_count):
weights = generate_weights(
NeuronCount(input=input_neuron_count, hidden=hidden_neuron_count))
assert sum(weights) == 0
import pytest
import numpy as np
from tests.stages.snn.utility import quarter
from snn_hfo_detection.stages.snn.weight_generation import generate_weights, POSSIBLE_ABSOLUTE_WEIGHTS
from snn_hfo_detection.stages.snn.concatenation import NeuronCount
TEST_NEURON_COUNT = [
NeuronCount(input=2, hidden=2),
NeuronCount(input=2, hidden=4),
NeuronCount(input=4, hidden=16),
NeuronCount(input=16, hidden=2)
]
@pytest.mark.parametrize('input_neuron_count, hidden_neuron_count',
TEST_NEURON_COUNT)
def test_weights_sum_to_zero(input_neuron_count, hidden_neuron_count):
weights = generate_weights(
NeuronCount(input=input_neuron_count, hidden=hidden_neuron_count))
assert sum(weights) == 0
@pytest.mark.parametrize('input_neuron_count, hidden_neuron_count',
TEST_NEURON_COUNT)
def test_weights_are_in_range(input_neuron_count, hidden_neuron_count):
weights = generate_weights(
NeuronCount(input=input_neuron_count, hidden=hidden_neuron_count))
min_weight = min(POSSIBLE_ABSOLUTE_WEIGHTS)
max_weight = max(POSSIBLE_ABSOLUTE_WEIGHTS)
for weight in weights:
assert min_weight <= weight <= max_weight or -max_weight <= weight <= -min_weight