def Experiment3():
mini_batch = 64
keep_prob = 0.5
# create model
model = Neuralnet()
model.add_layer(512, 1024, keep_prob=keep_prob, activation="sigmoid")
model.add_layer(64, keep_prob=keep_prob, activation="sigmoid")
model.add_layer(Y.shape[1], keep_prob=keep_prob, activation="no")
# Compile model
no_iterations = 1
learning_rate = 0.001
print("Learning-Rate:- ", learning_rate, " Batch_Size:- ", mini_batch)
losses, te_losses, out = model.train(training_f_X,
training_f_Y,
test_X,
test_Y,
learning_rate,
mini_batch,
no_iterations,
dropout=True)
title = "Exp3:- Batch size=" + str(mini_batch) + " Learning Rate=" + str(
learning_rate) + " with dropout"
plots.linear_plot([x for x in range(len(losses))], losses, te_losses,
"Iterations", "Losses", title, 5)
plt.savefig("../" + title + ".png")
np.savetxt("../3-64.csv", np.array(out), delimiter="|")
def adam():
mini_batch = 64
keep_prob = 1
# create model
model = Neuralnet()
model.add_layer(512, 1024, keep_prob=keep_prob, activation="sigmoid")
model.add_layer(64, keep_prob=keep_prob, activation="sigmoid")
model.add_layer(Y.shape[1], keep_prob=keep_prob, activation="no")
# Compile model
no_iterations = 100
learning_rate = 0.01
print("Learning-Rate:- ", learning_rate, " Batch_Size:- ", mini_batch)
losses, te_losses, out = model.train(training_f_X,
training_f_Y,
test_X,
test_Y,
learning_rate,
mini_batch,
no_iterations,
opt="adam")
title = "Adam:- Batch size=" + str(mini_batch) + " Learning Rate=" + str(
learning_rate)
plots.linear_plot([x for x in range(len(losses))], losses, te_losses,
"Iterations", "Losses", title, 1)
np.savetxt("../adam-" + ".csv", np.array(out), delimiter="|")
plt.savefig("../" + title + ".png")
def Experiment2():
mini_batchs = [32, 64, 128]
i = 0
for mini_batch in mini_batchs:
i += 1
keep_prob = 1
# create model
model = Neuralnet()
model.add_layer(512, 1024, keep_prob=keep_prob, activation="sigmoid")
model.add_layer(64, keep_prob=keep_prob, activation="sigmoid")
model.add_layer(Y.shape[1], keep_prob=keep_prob, activation="no")
# Compile model
no_iterations = 1000
learning_rate = 0.01
print("Learning-Rate:- ", learning_rate, " Batch_Size:- ", mini_batch)
losses, te_losses, out = model.train(training_f_X, training_f_Y,
test_X, test_Y, learning_rate,
mini_batch, no_iterations)
print(losses)
title = "Exp2:- Batch size=" + str(
mini_batch) + " Learning Rate=" + str(learning_rate)
plots.linear_plot([x for x in range(len(losses))], losses, te_losses,
"Iterations", "Losses", title, 1 + i)
plt.savefig("../" + title + ".png")
np.savetxt("../2-" + str(i) + ".csv", np.array(out), delimiter="|")
import pandas as pd
import numpy as np
import plots
from matplotlib import pyplot as plt
data = pd.read_csv('4-1.csv', sep="|", header=None)
iter = data.iloc[:, 0].values
tr_losses = data.iloc[:, 1].values
te_losses = data.iloc[:, 2].values
# iter = np.loadtxt( '2.csv' , delimiter='|', usecols=(0) )
title = "Exp4:- Batch size=" + str(64) + " Learning Rate=" + str(0.005)
plots.linear_plot(iter, tr_losses, te_losses, "Iterations", "Losses", title,
1 + 1)
plt.savefig(title + ".png")
plt.show()
