def p1_plots(q, dataSet):
fin = open("{}.pkl".format(dataSet), "rb")
train, test = pickle.load(fin)
X_tr, y_tr = train
X_te, y_te = test
X_te_ones = np.array([X_te[i] for i in range(len(X_te)) if y_te[i] == 1])
X_te_nils = np.array([X_te[i] for i in range(len(X_te)) if y_te[i] == 0])
X_tr_ones = np.array([X_tr[i] for i in range(len(X_tr)) if y_tr[i] == 1])
X_tr_nils = np.array([X_tr[i] for i in range(len(X_tr)) if y_tr[i] == 0])
k = []
if (q == 1):
k = [1, 2, 3, 4, 5]
elif (q == 2):
k = [1, 2, 3, 4, None]
elif (q == 3):
k = ["linear", "rbf", "poly"]
for i in k:
fig = plt.figure()
ax = fig.add_subplot(111)
if (q == 1):
ax.set_title("n_neighbors = %d" % (i))
elif (q == 2):
if (i is not None):
ax.set_title("max_depth = %d" % (i))
else:
ax.set_title("max_depth = None")
elif (q == 3):
ax.set_title("kernel = %s" % (i))
ax.scatter(X_tr_ones[:, 0], X_tr_ones[:, 1], c='b')
ax.scatter(X_tr_nils[:, 0], X_tr_nils[:, 1], c='r')
ax.scatter(X_te_ones[:, 0], X_te_ones[:, 1], c='b', marker="*")
ax.scatter(X_te_nils[:, 0], X_te_nils[:, 1], c='r', marker="*")
if (q == 1):
clf = KNeighborsClassifier(n_neighbors=i)
elif (q == 2):
clf = DecisionTreeClassifier(criterion='entropy', max_depth=i)
elif (q == 3):
clf = SVC(kernel=i)
clf.fit(X_tr, y_tr)
y_tr_pred = clf.predict(X_tr)
y_te_pred = clf.predict(X_te)
tr_loss = len(
[y_tr[i]
for i in range(len(y_tr)) if y_tr[i] != y_tr_pred[i]]) / len(y_tr)
te_loss = len(
[y_te[i]
for i in range(len(y_te)) if y_te[i] != y_te_pred[i]]) / len(y_te)
ax.set_xlabel("Tr Loss: {} \n Te Loss: {}".format(tr_loss, te_loss))
x_min, x_max, y_min, y_max = hlp.get_bounds(X_tr)
hlp.plot_decision_boundary(ax,
clf,
x_min=x_min,
x_max=x_max,
y_min=y_min,
y_max=y_max)
#plt.savefig("knn_plot_n%d.pdf" % (n))
# or if that takes too long: plot_decision_boundary(ax, clf, res=.1)
plt.show()
def decision_tree(pickle_file):
fin = open(pickle_file, "rb")
train, test = pickle.load(fin)
X_tr, y_tr = train
X_te, y_te = test
n = [1,2,3,4,0]
figure = plt.figure()
for spot in range(len(n)):
if n[spot] != 0:
dtc = DecisionTreeClassifier(criterion = "entropy", max_depth=n[spot])
else:
dtc = DecisionTreeClassifier(criterion = "entropy")
dtc.fit(X_tr, y_tr)
predicted_tr = (dtc.predict(X_tr))
predicted_te = (dtc.predict(X_te))
axis = figure.add_subplot(1,5,spot+1)
xtr_1 = []
xtr_2 = []
for pair in X_tr:
xtr_1.append(pair[0])
xtr_2.append(pair[1])
xte_1 = []
xte_2 = []
for pair in X_te:
xte_1.append(pair[0])
xte_2.append(pair[1])
colors = ListedColormap(['#FF0000', '#0000FF'])
axis.scatter(xtr_1,xtr_2, c = y_tr, cmap = colors, edgecolors = 'k')
axis.scatter(xte_1,xte_2, marker="*", c = y_te, cmap = colors, edgecolors = 'k')
x1min, x1max, x2min, x2max = helpers.get_bounds(X_tr)
helpers.plot_decision_boundary(axis, dtc, x1min, x1max, x2min, x2max)
if n[spot] != 0:
axis.set_title("max_depth = " + str(n[spot]))
else:
axis.set_title("max_depth = none")
tr_loss = round(zero_one_loss(y_tr,predicted_tr),2)
te_loss = round(zero_one_loss(y_te,predicted_te),2)
axis.set_xlabel("Tr loss: " + str(tr_loss)+"\n Te loss: " + str(te_loss))
plt.show()
def svm(pickle_file):
fin = open(pickle_file, "rb")
train, test = pickle.load(fin)
X_tr, y_tr = train
X_te, y_te = test
n = ["linear","rbf","poly"]
figure = plt.figure()
for spot in range(len(n)):
svc = SVC(kernel = n[spot])
svc.fit(X_tr, y_tr)
predicted_tr = (svc.predict(X_tr))
predicted_te = (svc.predict(X_te))
axis = figure.add_subplot(1,5,spot+1)
xtr_1 = []
xtr_2 = []
for pair in X_tr:
xtr_1.append(pair[0])
xtr_2.append(pair[1])
xte_1 = []
xte_2 = []
for pair in X_te:
xte_1.append(pair[0])
xte_2.append(pair[1])
colors = ListedColormap(['#FF0000', '#0000FF'])
axis.scatter(xtr_1,xtr_2, c = y_tr, cmap = colors, edgecolors = 'k')
axis.scatter(xte_1,xte_2, marker="*", c = y_te, cmap = colors, edgecolors = 'k')
x1min, x1max, x2min, x2max = helpers.get_bounds(X_tr)
helpers.plot_decision_boundary(axis, svc, x1min, x1max, x2min, x2max)
if n[spot] != 0:
axis.set_title("kernel = " + str(n[spot]))
else:
axis.set_title("kernel = none")
tr_loss = round(zero_one_loss(y_tr,predicted_tr),2)
te_loss = round(zero_one_loss(y_te,predicted_te),2)
axis.set_xlabel("Tr loss: " + str(tr_loss)+"\n Te loss: " + str(te_loss))
plt.show()
def kNN(pickle_file):
fin = open(pickle_file, "rb")
train, test = pickle.load(fin)
X_tr, y_tr = train
X_te, y_te = test
n = [1,3,5,7,9]
figure = plt.figure()
for spot in range(len(n)):
knc = KNeighborsClassifier(n_neighbors = n[spot])
knc.fit(X_tr, y_tr)
predicted_tr = (knc.predict(X_tr))
predicted_te = (knc.predict(X_te))
axis = figure.add_subplot(1,5,spot+1)
xtr_1 = []
xtr_2 = []
for pair in X_tr:
xtr_1.append(pair[0])
xtr_2.append(pair[1])
xte_1 = []
xte_2 = []
for pair in X_te:
xte_1.append(pair[0])
xte_2.append(pair[1])
colors = ListedColormap(['#FF0000', '#0000FF'])
axis.scatter(xtr_1,xtr_2, c = y_tr, cmap = colors, edgecolors = 'k')
axis.scatter(xte_1,xte_2, marker="*", c = y_te, cmap = colors, edgecolors = 'k')
x1min, x1max, x2min, x2max = helpers.get_bounds(X_tr)
helpers.plot_decision_boundary(axis, knc, x1min, x1max, x2min, x2max)
axis.set_title("n_neighbors = " + str(n[spot]))
tr_loss = round(zero_one_loss(y_tr,predicted_tr),2)
te_loss = round(zero_one_loss(y_te,predicted_te),2)
axis.set_xlabel("Tr loss: " + str(tr_loss)+"\n Te loss: " + str(te_loss))
plt.show()
lr=0.01
), # increase learning rate from 0.001 to 0.01 for faster learning
metrics=['accuracy'])
# Fit the model
history = model.fit(X_train, y_train, epochs=25)
# Evaluate our model on the test set
loss, accuracy = model.evaluate(X_test, y_test)
print(f"Model loss on the test set: {loss}")
print(f"Model accuracy on the test set: {100*accuracy:.2f}%")
# Plot the decision boundaries for the training and test sets
plt.figure(figsize=(12, 6))
plt.subplot(1, 2, 1)
plt.title("Train")
plot_decision_boundary(model, X=X_train, y=y_train)
plt.subplot(1, 2, 2)
plt.title("Test")
plot_decision_boundary(model, X=X_test, y=y_test)
plt.show()
# You can access the information in the history variable using the .history attribute
pd.DataFrame(history.history)
# Plot the loss curves
pd.DataFrame(history.history).plot()
plt.title("Model training curves")
model.save('models/my_model.h5')