The audit answers indicate r2 scores which deviate significantly from the values being independently achieved by various students doing the task. The MSE / MAE values are the same, whilst all other metrics are the same, suggesting that the underlying dataset is not the issue, but perhaps the way in which the training & test sets are being created have changed since the audit answers were last revised. Otherwise perhaps some sort of platform / system / architecture dependency...?
Users
Students following the AI specialization, grit:lab, Åland
Severity
(❗️minor)
Type
(🗂️ documentation)
To Reproduce
Steps to reproduce the behavior:
Jupyter Lab script
Exercise 3: Regression
print("\nExercise 3\n")
Fetch the dataset
housing = fetch_california_housing()
X, y = housing['data'], housing['target']
Split the dataset into training and test sets
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.1, shuffle=True, random_state=13)
print("r2 on the train set: ", r2_train)
print("MAE on the train set: ", mae_train)
print("MSE on the train set: ", mse_train)
print()
print("r2 on the test set: ", r2_test)
print("MAE on the test set: ", mae_test)
print("MSE on the test set: ", mse_test)
Workarounds
No workarounds, but if the audit template differs from the given results, it will be up to the auditor to judge pass / fail if the correct approach has been followed.
Expected behavior
The resulting r2_scores:
r2 on the train set: 0.6079874818809448
r2 on the test set: 0.5903435927516577
Attachments
N/A
Desktop (please complete the following information):
OS: Mac OS, Sonoma 14.3.1
Browser: N/A
Version: N/A
Smartphone (please complete the following information):
Describe the bug
The audit answers indicate r2 scores which deviate significantly from the values being independently achieved by various students doing the task. The MSE / MAE values are the same, whilst all other metrics are the same, suggesting that the underlying dataset is not the issue, but perhaps the way in which the training & test sets are being created have changed since the audit answers were last revised. Otherwise perhaps some sort of platform / system / architecture dependency...?
Users
Students following the AI specialization, grit:lab, Åland
Severity
(❗️minor)
Type
(🗂️ documentation)
To Reproduce
Steps to reproduce the behavior:
Jupyter Lab script
Exercise 3: Regression
print("\nExercise 3\n")
Fetch the dataset
housing = fetch_california_housing() X, y = housing['data'], housing['target']
Split the dataset into training and test sets
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.1, shuffle=True, random_state=13)
Define and configure the pipeline
pipeline = [('imputer', SimpleImputer(strategy='median')), ('scaler', StandardScaler()), ('lr', LinearRegression())] pipe = Pipeline(pipeline)
Fit the pipeline to the training data
pipe.fit(X_train, y_train)
Question 1: Predictions on Train and Test Sets
print("\nQuestion 1")
Predicting on the train set and test set
y_train_pred = pipe.predict(X_train) y_test_pred = pipe.predict(X_test)
Output the first 10 predicted values for both train and test sets
print("\n10 first values Train\n") print(y_train_pred[:10]) print("\n10 first values Test\n") print(y_test_pred[:10])
Question 2: Compute R2, MSE, and MAE
print("\n\nQuestion 2\n")
Compute R2, Mean Square Error, and Mean Absolute Error on the train set
r2_train = r2_score(y_train, y_train_pred) mse_train = mean_squared_error(y_train, y_train_pred) mae_train = mean_absolute_error(y_train, y_train_pred)
Compute R2, Mean Square Error, and Mean Absolute Error on the test set
r2_test = r2_score(y_test, y_test_pred) mse_test = mean_squared_error(y_test, y_test_pred) mae_test = mean_absolute_error(y_test, y_test_pred)
Print the results
print("r2 on the train set: ", r2_train) print("MAE on the train set: ", mae_train) print("MSE on the train set: ", mse_train) print() print("r2 on the test set: ", r2_test) print("MAE on the test set: ", mae_test) print("MSE on the test set: ", mse_test)
Workarounds
No workarounds, but if the audit template differs from the given results, it will be up to the auditor to judge pass / fail if the correct approach has been followed.
Expected behavior
The resulting r2_scores:
Attachments
N/A
Desktop (please complete the following information):
Smartphone (please complete the following information):
N/A
Additional context
N/A