def update_graph(dt, var, ntr, nte, mod, deg, alp):
np.random.seed(100)
x_train, y_train, x_test, y_test = 0, 0, 0, 0
if dt in ['sin', 'cos', 'exp', 'log']:
if dt == 'sin':
x_data = random.uniform(0, 1, ntr)
x_train = x_data.reshape(-1, 1) # create a matrix from the list
y_train = np.sin(2 * np.pi * x_data) + random.normal(
0, var, x_data.shape)
x_data_test = random.uniform(0, 1, nte)
x_test = x_data_test.reshape(-1, 1)
y_test = np.sin(2 * np.pi * x_data_test) + random.normal(
0, var, x_data_test.shape)
if dt == 'cos':
x_data = random.uniform(0, 1, ntr)
x_train = x_data.reshape(-1, 1) # create a matrix from the list
y_train = np.cos(2 * np.pi * x_data) + random.normal(
0, var, x_data.shape)
x_data_test = random.uniform(0, 1, nte)
x_test = x_data_test.reshape(-1, 1)
y_test = np.cos(2 * np.pi * x_data_test) + random.normal(
0, var, x_data_test.shape)
if dt == 'log':
x_data = random.uniform(0, 1, ntr)
x_train = x_data.reshape(-1, 1) # create a matrix from the list
y_train = np.log(x_data) + random.normal(0, var, x_data.shape)
x_data_test = random.uniform(0, 1, nte)
x_test = x_data_test.reshape(-1, 1)
y_test = np.log(x_data_test) + random.normal(
0, var, x_data_test.shape)
if dt == 'exp':
x_data = random.uniform(0, 1, ntr)
x_train = x_data.reshape(-1, 1) # create a matrix from the list
y_train = np.exp(x_data) + random.normal(0, var, x_data.shape)
x_data_test = random.uniform(0, 1, nte)
x_test = x_data_test.reshape(-1, 1)
y_test = np.exp(x_data_test) + random.normal(
0, var, x_data_test.shape)
X_train = PolynomialFeatures(degree=deg,
include_bias=True).fit_transform(x_train)
X_test = PolynomialFeatures(degree=deg,
include_bias=True).fit_transform(x_test)
if mod == 'lasso':
model = Lasso(alpha=alp, normalize=True)
elif mod == 'ridge':
model = Ridge(alpha=alp, normalize=True)
else:
model = LinearRegression(normalize=True)
model.fit(X_train, y_train)
Xmax = X_train
if X_train.max() < X_test.max():
Xmax = X_test
Xmin = X_train
if X_train.min() > X_test.min():
Xmin = X_test
x_range = np.linspace(Xmin.min(), Xmax.max(), 1000,
endpoint=True).reshape(-1, 1)
X_range_poly = PolynomialFeatures(degree=deg,
include_bias=True).fit_transform(x_range)
y_range = model.predict(X_range_poly)
test_score = model.score(X_test, y_test)
test_error = mean_squared_error(y_test, model.predict(X_test))
train_score = model.score(X_train, y_train)
train_error = mean_squared_error(y_train, model.predict(X_train))
eq = format_coefs(model.coef_.round(2))
fig = go.Figure([
go.Scatter(x=x_train.squeeze(),
y=y_train,
name='train',
mode='markers'),
go.Scatter(x=x_test.squeeze(), y=y_test, name='test', mode='markers'),
go.Scatter(x=x_range.squeeze(), y=y_range, name=eq),
])
fig.update_layout(
# title= str(dataset) + " Linear Regression Plot",
xaxis_title='X',
yaxis_title="Dependent Variable",
legend_title="Legend",
title=f"test_MSE: {test_error:.3f}, train_MSE: {train_error:.3f}",
legend=dict(orientation='h', ),
margin=dict(l=25, r=15, t=50, b=50),
hovermode='closest',
font=dict(family="Courier New, monospace",
size=12,
color="RebeccaPurple"))
return fig
