def experiment6():
k = 1
n_features_list = np.arange(0, 30, 2) + 2
n_data = 10**4
n_targets = 1000
results = []
for n_features in n_features_list:
data = np.random.uniform(0, 1, n_data * n_features).reshape(
(n_data, n_features))
targets = np.random.uniform(0, 1, n_targets * n_features).reshape(
(n_targets, n_features))
model = Trees.VPTree(data, 2, similarity_metric='euclidean')
number_of_comparisons = experiment_number_of_comparisons_model(
targets, data, model, k, 6, 'vp_tree_2')
results.append(number_of_comparisons)
plt.clf()
plt.boxplot(results)
plt.xticks(np.arange(len(n_features_list)) + 1, n_features_list)
plt.title(
'Number of comparisions in vp tree, for different dimensionalities\nNumber of data points: {0}\nNumber of trials: {1}'
.format(n_data, n_targets))
plt.ylabel('number of comparisons')
plt.xlabel('number of dimensions')
plt.margins(0.2, 0.2)
plt.savefig('simulation6_number_of_comparisons.jpg', bbox_inches='tight')
def experiment1():
print('Running simulation 1')
n_data = 10**4
n_features = 30
k = 1
data = np.random.uniform(0, 1, n_data * n_features).reshape(
(n_data, n_features))
n_targets = 1000
targets = np.random.uniform(0, 1, n_targets * n_features).reshape(
(n_targets, n_features))
model_names = [
'kd-tree', 'vp-tree-2', 'vp-tree-10', 'vp-tree-100', 'pp-100-100',
'pp-1000-1000'
]
results = []
for i in range(2):
if i == 0:
model = Trees.KDTree(data, similarity_metric='euclidean')
elif i == 1:
model = Trees.VPTree(data, 2, similarity_metric='euclidean')
elif i == 2:
model = Trees.VPTree(data, 10, similarity_metric='euclidean')
elif i == 3:
model = Trees.VPTree(data, 100, similarity_metric='euclidean')
elif i == 4:
model = NewDataStructures.PavlosIndexingSystem(
data, 100, 100, similarity_metric='euclidean')
else:
model = NewDataStructures.PavlosIndexingSystem(
data, 1000, 1000, similarity_metric='euclidean')
number_of_comparisons = experiment1_model(targets, data, model, k,
model_names[i])
results.append(number_of_comparisons)
plt.clf()
plt.boxplot(results)
plt.margins(0.2, 0.2)
plt.xticks(np.arange(len(model_names)) + 1,
model_names,
rotation='vertical')
plt.title(
'Number of comparisions per model\nNumber of data points: {0}\nDimensionality: {1}\nNumber of trials: {2}\nk={3}'
.format(n_data, n_features, n_targets, k))
plt.ylabel('number of comparisons')
plt.savefig('simulation1_number_of_comparisons.jpg', bbox_inches='tight')
