def test_bbclassifier():
c = BBClassifier(num_epochs=3,
use_normal_dist_bases=True,
use_evenly_spaced_periods=True,
random_state=rand_seed)
train_data = np.array([
0.0, 1.0, 0.0, 1.0, 0.0, 1.0, 0.0, 1.0, 0.0, 1.0, 0.0, 1.0, 0.0, 1.0,
0.0, 1.0, 0.0, 1.0, 0.0, 1.0, 0.0, 1.0, 0.0, 1.0, 0.0, 1.0, 0.0, 1.0,
0.0, 1.0
]).reshape(-1, 1)
train_labels = np.array([
0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1,
0, 1, 0, 1, 0, 1
]).reshape(-1)
c.fit(train_data, train_labels)
preds = c.predict([[0.0]])
actual_preds = np.array(preds)
expect_preds = np.array([0])
np.testing.assert_array_equal(actual_preds, expect_preds)
preds = c.predict([[1.0]])
actual_preds = np.array(preds)
expect_preds = np.array([1])
np.testing.assert_array_equal(actual_preds, expect_preds)
# h = .05 # step size in the mesh
h = .02 # step size in the mesh
num_samples = 500
rand_seed = 42
color_maps = ['Reds', 'Blues', 'Greens', 'Oranges', 'Purples', 'Greys']
color_list = ['r', 'b', 'g', 'orange', 'purple', 'grey']
names = [
"BBClassifier", "Nearest Neighbors", "Decision Tree", "Neural Net",
"Naive Bayes"
]
classifiers = [
BBClassifier(num_epochs=3,
use_normal_dist_bases=True,
use_evenly_spaced_periods=True,
random_state=rand_seed),
KNeighborsClassifier(3),
DecisionTreeClassifier(max_depth=5, random_state=rand_seed),
MLPClassifier(alpha=1,
max_iter=1000,
hidden_layer_sizes=(100, 100),
random_state=rand_seed),
GaussianNB()
]
noise = 0.1
X, y = make_classification(n_samples=num_samples,
n_features=2,
n_redundant=0,
# ==============================================================================
# sklearn_style_classifier.py
# ==============================================================================
from brainblocks.tools import BBClassifier
from sklearn.datasets import make_classification
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import MinMaxScaler
import time
import numpy as np
h = .02 # step size in the mesh
num_samples = 500
rand_seed = 42
noise = 0.1
clf = BBClassifier(num_epochs=3, random_state=rand_seed)
X, y = make_classification(n_samples=num_samples,
n_features=2,
n_redundant=0,
n_informative=2,
random_state=rand_seed,
n_clusters_per_class=1,
n_classes=2)
rng = np.random.RandomState(rand_seed)
X += 2 * rng.uniform(low=0.0, high=noise, size=X.shape)
# scale the data
X = MinMaxScaler().fit_transform(X)
def evaluate(datasets, classifier_configs):
# Evaluation Configuration
#
# h = .02
h = 0.05
x_boundary = 0.5
y_boundary = 0.5
# x_boundary = 6.5
# y_boundary = 6.5
# x_boundary = 2.5
# y_boundary = 2.5
# j_boundary = 12.0
# y_boundary = 12.0
# iterate over datasets
results_per_dataset = []
for ds_cnt, ds in enumerate(datasets):
print("Data Set %d" % ds_cnt)
# preprocess dataset, split into training and test part
X, y, X_plot = ds
X = MinMaxScaler().fit_transform(X)
X_plot = MinMaxScaler().fit_transform(X_plot)
# uniform mesh
# x_min, x_max = X[:, 0].min() - x_boundary, X[:, 0].max() + x_boundary
# y_min, y_max = X[:, 1].min() - y_boundary, X[:, 1].max() + y_boundary
# xx, yy = np.meshgrid(np.arange(x_min, x_max, h),
# np.arange(y_min, y_max, h))
X_train, X_test, y_train, y_test, X_plot_train, X_plot_test = \
train_test_split(X, y, X_plot, test_size=.1, random_state=42)
# all data in a tuple for comparison across classifiers and for plotting
# split_data = (X, y, xx, yy, X_train, X_test, y_train, y_test)
split_data = (X, y, X_plot, X_train, X_test, y_train, y_test,
X_plot_train, X_plot_test)
# collect results for each classifier experiment on this data
results_per_classifier = []
# iterate over classifiers
for name, bb_config in classifier_configs:
num_input_dims = X_train.shape[1]
bb_config['num_input_dims'] = num_input_dims
# clf = KNeighborsClassifier(3)
clf = BBClassifier(**bb_config)
print("Train %s" % name)
# train the model
clf.fit(X_train, y_train)
# compute accuracy on test values
score = clf.score(X_test, y_test)
results_per_classifier.append((name, score))
del clf
data_result = {'data': split_data, 'results': results_per_classifier}
results_per_dataset.append(data_result)
return results_per_dataset