def train_ignore_features():
data_limit = 10
model_name = 'first_linear_then_more_GraphConvs_then_linear'
predict_type = 'y_combine_all_percent'
train_d, train_f, val_d, val_f, word_to_ixs = data(data_limit)
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
print('Number of features', NODE_FEATURES_NUM)
x_list = list(range(NODE_FEATURES_NUM))
auc_list = []
for i in x_list:
my_models = MyModels(word_to_ixs, ignore_columns=[i])
net = my_models.my_models[model_name]
criterion = nn.CrossEntropyLoss(
weight=torch.tensor([1.0, 6.0], device=device))
training_h, validation_h, whole_training_h = run(net,
train_d,
val_d,
device=device,
criterion=criterion,
batch_size=10,
epochs=2,
model_name=model_name,
save=False)
thresholds, aucs = calculate_metrics(val_d,
net,
print_model_name=None,
do_plot=False,
save=False)
auc_list.append(aucs[predict_type])
plt.figure()
plt.bar(x_list, auc_list)
plt.show()
import main as mn
import sqlite3 as sql
con = sql.connect("Pyth")
usr = input("Username: "******"Password: "******"insert into data values ('" + usr + "','" + x["hash"] + "','" +
str(x["num"]) + "')")
con.commit()
import altair as alt
import main
import pandas as pd
source = main.data()
chart = alt.Chart(source).mark_bar().encode(x='id', y="loc")
# The highlight will be set on the result of a conditional statement
chart_1 = alt.Chart(source).mark_circle(size=60).encode(
x='id',
y='CC',
color='file_name',
#tooltip=['Origin', 'Horsepower', 'Miles_per_Gallon']
).interactive()
chart_2 = alt.Chart(source).mark_line().encode(x='id', y='CC')
chart.save('chart.html')
false_pos = []
right = []
A = 1
B = 1
C = 1
D = 1
E = 1
correct_attempts = 0
THRESHOLD = 500
with open("data.txt", "r") as f:
for line in f.readlines():
pos = line.find(":")
userdata = ""
exec("userdata = " + line[pos + 1:-1])
todo.append([line[0:pos], main.data(userdata)])
while percentage < 70 or not false_pos:
wrong = []
right = []
false_pos = []
correct_attempts = 0
for item in todo:
score = user.checkTest(item[1], A, B, C, D, E)
if (score >= THRESHOLD and item[0] == "joe"):
correct_attempts += 1
wrong.append(item)
elif (score < THRESHOLD and item[0] != "joe"):
false_pos.append(item)
wrong.append(item)
else:
import numpy as np
import main, pickle
#load the data from the main script
norm = main.data()[0]
test_data = main.data()[3]
test_labels = main.data()[4]
total = main.data()[6]
hot_label = main.data()[7]
#load the parameters from the main script
net = main.network()[0]
#activation function
def f(z):
if z > 100.0:
return 1.0
elif z < -100.0:
return 0.0
else:
return 1.0 / (1.0 + np.exp(-z))
x = [] #input to each neuron
y = [] #activation of each neuron
for layer in range(len(net)):
x.append(np.array([0.0] * net[layer]))
y.append(np.array([0.0] * net[layer]))
import random, sys, pickle
import numpy as np
import main, test
#load the data from the main script
norm = main.data()[0]
training_data = main.data()[1]
training_labels = main.data()[2]
check_test = main.data()[5]
hot_label = main.data()[7]
#load the parameters from the main script
net = main.network()[0]
epochs = main.network()[1]
number_examples = main.network()[2]
mini_batch_size = main.network()[3]
suffle_training = main.network()[4]
learning_rate = main.network()[5]
momentum_rate = main.network()[6]
tau = main.network()[7]
period = main.network()[8]
#activation function
def f(z):
if z > 100.0:
return 1.0
elif z < -100.0:
return 0.0
else:
return 1.0 / (1.0 + np.exp(-z))