Simple neural network with scikit-learn in ML Python - Practice Problems & Coding Challenges

Challenge - 5 Problems

❓ Predict Output

intermediate

Output of training a simple neural network

What will be the printed output of the training accuracy after fitting this neural network on the given data?

ML Python

from sklearn.neural_network import MLPClassifier
from sklearn.datasets import load_iris
from sklearn.model_selection import train_test_split

X, y = load_iris(return_X_y=True)
X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=42)

model = MLPClassifier(hidden_layer_sizes=(5,), max_iter=500, random_state=1)
model.fit(X_train, y_train)

print(f"Training accuracy: {model.score(X_train, y_train):.2f}")

A Training accuracy: 1.00

B Training accuracy: 0.50

C Training accuracy: 0.75

D Training accuracy: 0.00

💡 Hint

The model is trained with enough iterations on a simple dataset, so it should fit well.

The MLPClassifier with 5 hidden neurons and 500 iterations on the Iris training set achieves perfect or near-perfect training accuracy, so the printed accuracy is 1.00.

❓ Model Choice

intermediate

Choosing the correct hidden layer size

Which hidden_layer_sizes parameter will create a neural network with two hidden layers, the first with 10 neurons and the second with 5 neurons?

A (5, 10)

B (10, 5)

C 10

D (15,)

💡 Hint

The tuple defines the number of neurons per hidden layer in order.

The hidden_layer_sizes parameter takes a tuple where each number represents the neurons in each hidden layer. (10, 5) means first layer 10 neurons, second layer 5 neurons.

❓ Hyperparameter

advanced

Effect of max_iter on training

What will happen if max_iter is set too low (e.g., max_iter=1) when training an MLPClassifier on a dataset?

A The model will overfit and training accuracy will be very high.

B The model will raise a runtime error and stop.

C The model will train normally with no effect on accuracy.

D The model will likely underfit and training accuracy will be low.

💡 Hint

Training too few iterations means the model does not learn enough.

With max_iter=1, the model barely trains, so it underfits and training accuracy remains low.

❓ Metrics

advanced

Interpreting loss_curve_ attribute

What does the loss_curve_ attribute of a trained MLPClassifier represent?

A A list of loss values at each iteration during training.

B The final loss value after training completes.

C The weights of the neural network layers.

D The accuracy values on the test set during training.

💡 Hint

Think about what 'curve' means in this context.

loss_curve_ stores the loss value at each iteration, showing how the model improved during training.

🔧 Debug

expert

Identifying error in neural network code

What error will this code raise when run? from sklearn.neural_network import MLPClassifier model = MLPClassifier(hidden_layer_sizes=5, max_iter=200) model.fit([[0,0],[1,1]], [0,1])

ML Python

from sklearn.neural_network import MLPClassifier

model = MLPClassifier(hidden_layer_sizes=5, max_iter=200)
model.fit([[0,0],[1,1]], [0,1])

A ValueError: hidden_layer_sizes must be a tuple

B TypeError: 'int' object is not iterable

C No error, code runs successfully

D AttributeError: 'MLPClassifier' object has no attribute 'fit'

💡 Hint

scikit-learn accepts integers for hidden_layer_sizes, converting to a single-layer tuple.

MLPClassifier accepts an integer for hidden_layer_sizes, which is interpreted as a single hidden layer with that many neurons. No error occurs.