def Tensorflow_neural_network(X, y, Xt, yt):
"""
Classification
Neural Networks
Tensorflows module
"""
print("-------------------")
yp = fns.tensorflow_NNWsolver(X, y, Xt, yt)
# print(yp)
a = fns.assert_binary_accuracy(yt, yp)
return f"Tensorflow NN accuracy: {100*a:.0f} %"
def Sklearn_sgd_classifier(X, y, Xt, yt):
"""
Classification
Logistic Regression
Scikit-learns module: SGDClassifier
"""
print("-------------------")
solution = sklearn.linear_model.SGDClassifier(eta0=0.01,
max_iter=100).fit(X, y)
yp = solution.predict(X)
# print(yp)
a = fns.assert_binary_accuracy(y, yp)
return f"Sklearn\'s SGDClassifier accuracy: {100*a:.0f} %"
def declare_results(self, ytrue, ypred, round=True):
"""Declare the results, assign the accuracy"""
if round:
ypred[np.where(ypred>=0.5)] = 1
ypred[np.where(ypred<0.5)] = 0
for i in range(len(ytrue)):
if i%20==0:
print(f"Output:\t{ypred[i]}\tTrue:\t{ytrue[i]}" + (\
" Correct!" if ypred[i]==ytrue[i] else " "))
else:
for i in range(len(ytrue)):
print(f"Output:\t{ypred[i]}\tTrue:\t{ytrue[i]}")
self.acc = fns.assert_binary_accuracy(ytrue, ypred)
def SGD_with_minibatches(X,
y,
Xt,
yt,
gamma,
max_iter,
batch_size,
verbose=False):
"""
Classification
Logistic Regression
Our own SGD with mini-batches
(see "./lib/logistic_regressionk.py")
"""
obj = lgr.StochasticGradientDescent(gamma,
max_iter,
batch_size,
verbose=verbose)
obj.fit(X, y)
yp = obj.predict(Xt)
# print(yp)
a = fns.assert_binary_accuracy(y, yp)
return f"SGD with mini-batches accuracy: {100*a:.0f} %"