def main():
plt.rcParams['figure.figsize'] = (7.0, 4.0) # set default size of plots
plt.rcParams['image.interpolation'] = 'nearest'
plt.rcParams['image.cmap'] = 'gray'
train_X, train_Y = load_dataset()
layers_dims = [train_X.shape[0], 5, 2, 1]
#parameters = model(train_X, train_Y, layers_dims, optimizer = "gd")
#parameters = model(train_X, train_Y, layers_dims, beta = 0.9, optimizer = "momentum")
parameters = model(train_X, train_Y, layers_dims, optimizer="adam")
# Predict
predictions = predict(train_X, train_Y, parameters)
# Plot decision boundary
#plt.title("Model with Gradient Descent optimization")
#plt.title("Model with Momentum optimization")
plt.title("Model with Adam optimization")
axes = plt.gca()
axes.set_xlim([-1.5, 2.5])
axes.set_ylim([-1, 1.5])
plot_decision_boundary(lambda x: predict_dec(parameters, x.T), train_X,
train_Y)
plt.show()
def main(optimizer='gd'):
# load and plot the data points
train_X, train_Y = load_dataset()
plt.show()
layers_dims = [train_X.shape[0], 5, 2, 1]
# try different optimization methods
if optimizer == 'gd':
# mini-batch gradient descent
parameters = model(train_X, train_Y, layers_dims, optimizer="gd")
plt.title("Model with Gradient Descent optimization")
elif optimizer == 'momentum':
# mini-batch gradient descent with momentum
parameters = model(train_X, train_Y, layers_dims, optimizer="momentum")
plt.title("Model with Momentum optimization")
elif optimizer == 'adam':
# mini-batch gradient descent with Adam
parameters = model(train_X, train_Y, layers_dims, optimizer="adam")
plt.title("Model with Adam optimization")
else:
print("No such optimization method named " + optimizer)
return
# Predict
predictions = predict(train_X, train_Y, parameters)
# Plot decision boundary
axes = plt.gca()
axes.set_xlim([-1.5, 2.5])
axes.set_ylim([-1, 1.5])
plot_decision_boundary(lambda x: predict_dec(parameters, x.T), train_X,
train_Y)
def train_adam():
# train 3-layer model
layers_dims = [train_X.shape[0], 5, 2, 1]
parameters = model(train_X, train_Y, layers_dims, optimizer="adam")
# Predict
predictions = predict(train_X, train_Y, parameters)
# Plot decision boundary
plt.title("Model with Adam optimization")
axes = plt.gca()
axes.set_xlim([-1.5, 2.5])
axes.set_ylim([-1, 1.5])
plot_decision_boundary(lambda x: predict_dec(parameters, x.T), train_X,
train_Y)
def mini_batch_with_gradient():
train_X, train_Y = load_dataset()
# train 3-layer model
layers_dims = [train_X.shape[0], 5, 2, 1]
parameters = model(train_X, train_Y, layers_dims, optimizer="gd")
# Predict
predictions = predict(train_X, train_Y, parameters)
# Plot decision boundary
plt.title("Model with Gradient Descent optimization")
axes = plt.gca()
axes.set_xlim([-1.5, 2.5])
axes.set_ylim([-1, 1.5])
plot_decision_boundary(lambda x: predict_dec(parameters, x.T), train_X,
train_Y)
def main():
train_X, train_Y = load_dataset()
layers_dims = [train_X.shape[0], 5, 2, 1]
print("training with mini-batch gd optimizer")
learned_parameters_gd = model(train_X,
train_Y,
layers_dims,
optimizer="gd")
predict(train_X, train_Y, learned_parameters_gd)
# Plot decision boundary
plt.title("model with Gradient Descent optimization")
axes = plt.gca()
axes.set_xlim([-1.5, 2.5])
axes.set_ylim([-1, 1.5])
plot_decision_boundary(lambda x: predict_dec(learned_parameters_gd, x.T),
train_X, train_Y)
print("training with momentum optimizer")
learned_parameters_momentum = model(train_X,
train_Y,
layers_dims,
beta=0.9,
optimizer="momentum")
predict(train_X, train_Y, learned_parameters_momentum)
# Plot decision boundary
plt.title("model with Momentum optimization")
axes = plt.gca()
axes.set_xlim([-1.5, 2.5])
axes.set_ylim([-1, 1.5])
plot_decision_boundary(
lambda x: predict_dec(learned_parameters_momentum, x.T), train_X,
train_Y)
print("training with adam optimizer")
learned_parameters_adam = model(train_X,
train_Y,
layers_dims,
optimizer="adam")
predict(train_X, train_Y, learned_parameters_adam)
# Plot decision boundary
plt.title("model with Adam optimization")
axes = plt.gca()
axes.set_xlim([-1.5, 2.5])
axes.set_ylim([-1, 1.5])
plot_decision_boundary(lambda x: predict_dec(learned_parameters_adam, x.T),
train_X, train_Y)
return None
return parameters
# train 3-layer model
layers_dims = [train_X.shape[0], 5, 2, 1]
parameters = model(train_X, train_Y, layers_dims, optimizer="gd")
# Predict
predictions = predict(train_X, train_Y, parameters)
# Plot decision boundary
plt.title("Model with Gradient Descent optimization")
axes = plt.gca()
axes.set_xlim([-1.5, 2.5])
axes.set_ylim([-1, 1.5])
plot_decision_boundary(lambda x: predict_dec(parameters, x.T), train_X,
train_Y)
# train 3-layer model
layers_dims = [train_X.shape[0], 5, 2, 1]
parameters = model(train_X,
train_Y,
layers_dims,
num_epochs=30000,
beta=0.9,
optimizer="momentum")
# Predict
predictions = predict(train_X, train_Y, parameters)
# Plot decision boundary
plt.title("Model with Momentum optimization")
# In[16]:
# train 3-layer model
layers_dims = [train_X.shape[0], 5, 2, 1]
parameters = model(train_X, train_Y, layers_dims, optimizer = "gd")
# Predict
predictions = predict(train_X, train_Y, parameters)
# Plot decision boundary
plt.title("Model with Gradient Descent optimization")
axes = plt.gca()
axes.set_xlim([-1.5,2.5])
axes.set_ylim([-1,1.5])
plot_decision_boundary(lambda x: predict_dec(parameters, x.T), train_X, train_Y)
# ### 5.2 - Mini-batch gradient descent with momentum
#
# Run the following code to see how the model does with momentum. Because this example is relatively simple, the gains from using momemtum are small; but for more complex problems you might see bigger gains.
# In[17]:
# train 3-layer model
layers_dims = [train_X.shape[0], 5, 2, 1]
parameters = model(train_X, train_Y, layers_dims, beta = 0.9, optimizer = "momentum")
# Predict
predictions = predict(train_X, train_Y, parameters)
t = t + 1
self.Update_Parameters_Adam(t, learning_rate, beta1, beta2, epsilon)
if print_cost and i % (10*self.iteration_unit) == 0:
print ("Cost after iteration %i: %f" %(i, cost))
self.Query(X, Y)
if i % self.iteration_unit == 0:
self.costs.append(cost)
print('finished!')
self.PlotCosts()
return self.costs
pass
train_X, train_Y = load_dataset()
plt.show()
a = DeepNeuralNetwork("Model without optimization", [train_X.shape[0], 10, 3, 1])
for i in range(1500):
a.Forward_Propagation(train_X)
a.Backward_Propagation(train_Y)
a.Update_Parameters(0.35)
if i % 50 == 0:
plt.title("Model without optimization")
axes = plt.gca()
axes.set_xlim([-1.5,2.5])
axes.set_ylim([-1,1.5])
plot_decision_boundary(lambda x: a.Predict(x.T), train_X, np.squeeze(train_Y))
plt.show()
a.Query(train_X, train_Y)
# Test model using **common gradient descent**
# In[27]:
layer_dims = [train_X.shape[0], 5, 2, 1]
params = model(train_X, train_Y, layer_dims, optimizer="gd", is_plot=True)
# In[28]:
predictions = opt_utils.predict(train_X, train_Y, params)
plt.title("Gradient Descent")
axes = plt.gca()
axes.set_xlim([-1.5, 2.5])
axes.set_ylim([-1, 1.5])
opt_utils.plot_decision_boundary(lambda x: opt_utils.predict_dec(params, x.T),
train_X, np.squeeze(train_Y))
# Test model using **momentum gradient descent**
# In[29]:
layer_dims = [train_X.shape[0], 5, 2, 1]
params = model(train_X,
train_Y,
layer_dims,
beta=0.9,
optimizer="momentum",
is_plot=True)
# In[30]:
layers_dims = [train_X.shape[0], 5, 2, 1]
parameters = model(train_X, train_Y, layers_dims, optimizer="momentum", is_plot=True)
preditions = opt_utils.predict(train_X, train_Y, parameters)
# 绘制分类图
plt.title("Model with Gradient Descent optimization")
axes = plt.gca()
axes.set_xlim([-1.5, 2.5])
axes.set_ylim([-1, 1.5])
opt_utils.plot_decision_boundary(lambda x: opt_utils.predict_dec(parameters, x.T), train_X, train_Y)
'''
#测试adam优化器
train_X, train_Y = opt_utils.load_dataset(is_plot=False)
layers_dims = [train_X.shape[0], 5, 2, 1]
parameters = model(train_X,
train_Y,
layers_dims,
optimizer="adam",
is_plot=True)
preditions = opt_utils.predict(train_X, train_Y, parameters)
# 绘制分类图
plt.title("Model with Gradient Descent optimization")
axes = plt.gca()
axes.set_xlim([-1.5, 2.5])
axes.set_ylim([-1, 1.5])
opt_utils.plot_decision_boundary(
lambda x: opt_utils.predict_dec(parameters, x.T), train_X, train_Y)
sep("Mini-batch gradient descent")
# train 3-layer model
layers_dims = [train_X.shape[0], 5, 2, 1]
parameters = model(train_X, train_Y, layers_dims, optimizer = "gd")
# Predict
predictions = predict(train_X, train_Y, parameters)
# Plot decision boundary
plt.title("Model with Gradient Descent optimization")
axes = plt.gca()
axes.set_xlim([-1.5,2.5])
axes.set_ylim([-1,1.5])
plot_decision_boundary(lambda x: predict_dec(parameters, x.T), train_X, np.reshape(train_Y,-1))
sep("with momentum")
# train 3-layer model
layers_dims = [train_X.shape[0], 5, 2, 1]
parameters = model(train_X, train_Y, layers_dims, beta = 0.9, optimizer = "momentum")
# Predict
predictions = predict(train_X, train_Y, parameters)
# Plot decision boundary
plt.title("Model with Momentum optimization")
axes = plt.gca()
axes.set_xlim([-1.5,2.5])
axes.set_ylim([-1,1.5])
plot_decision_boundary(lambda x: predict_dec(parameters, x.T), train_X, np.reshape(train_Y,-1))
return parameters
# train 3-layer model
layers_dims = [train_X.shape[0], 5, 2, 1]
parameters = model(train_X, train_Y, layers_dims, optimizer = "gd")
# Predict
predictions = predict(train_X, train_Y, parameters)
# Plot decision boundary
plt.title("Model with Gradient Descent optimization")
axes = plt.gca()
axes.set_xlim([-1.5,2.5])
axes.set_ylim([-1,1.5])
plot_decision_boundary(lambda x: predict_dec(parameters, x.T), train_X, np.squeeze(train_Y))
plt.show()
# train 3-layer model
layers_dims = [train_X.shape[0], 5, 2, 1]
parameters = model(train_X, train_Y, layers_dims, beta = 0.9, optimizer = "momentum")
# Predict
predictions = predict(train_X, train_Y, parameters)
# Plot decision boundary
plt.title("Model with Momentum optimization")
axes = plt.gca()
axes.set_xlim([-1.5,2.5])
axes.set_ylim([-1,1.5])
plot_decision_boundary(lambda x: predict_dec(parameters, x.T), train_X, np.squeeze(train_Y))