Machine Learning 304 (PT2)

What is the result of multiplying a 2x3 matrix by a 3x2 matrix?

• The result is a 1x2 matrix.

• The result is a 2x1 matrix.

• The result is a 2x2 matrix.

• The result is a 3x3 matrix.

How do you compute the transpose of a matrix?

• The transpose of a matrix is obtained by rotating it 90 degrees.

• The transpose of a matrix is found by multiplying it by its inverse.

• The transpose of a matrix is obtained by swapping its rows and columns.

• The transpose of a matrix is computed by adding its rows and columns.

What is the purpose of the learning rate in gradient descent?

• To adjust the weight initialization values

• The purpose of the learning rate in gradient descent is to determine the step size at each iteration while moving toward a minimum of the loss function.

• To increase the complexity of the model

• To reduce the number of iterations needed

Explain the difference between batch gradient descent and stochastic gradient descent.

• Batch gradient descent updates weights after every epoch, while stochastic gradient descent updates after multiple epochs.

• Batch gradient descent is faster than stochastic gradient descent for large datasets, while stochastic is slower.

• Batch gradient descent uses the entire dataset for each update, while stochastic gradient descent uses one data point at a time.

• Batch gradient descent uses random samples for updates, while stochastic uses the entire dataset.

What is the role of the activation function in a neural network?

• The activation function increases the network's speed.

• The activation function determines the number of layers in the network.

• The activation function allows the neural network to learn non-linear relationships.

• The activation function is used for data normalization.

Describe the architecture of a convolutional neural network (CNN).

• A CNN is primarily used for natural language processing tasks.

• A CNN architecture includes input layer, convolutional layers, activation functions, pooling layers, and fully connected layers.

• A CNN architecture consists of only input and output layers.

• A CNN includes only fully connected layers and no convolutional layers.

What is the main idea behind support vector machines (SVM)?

• SVM is used to reduce the dimensionality of data.

• The main idea of SVM is to cluster data points into groups.

• The main idea behind support vector machines (SVM) is to find the optimal hyperplane that maximizes the margin between different classes.

• SVM focuses on minimizing the distance between all data points.

How does regularization help prevent overfitting in machine learning models?

• Regularization helps prevent overfitting by adding a penalty for complexity, encouraging simpler models that generalize better.

• Regularization increases model complexity for better accuracy.

• Regularization only works with linear models.

• Regularization has no effect on model performance.

What is L1 regularization and how does it differ from L2 regularization?

• L1 regularization squares the coefficients, while L2 regularization uses absolute values.

• L1 regularization leads to larger coefficients, while L2 regularization leads to sparsity.

• L1 regularization adds the absolute value of coefficients as a penalty, leading to sparsity; L2 regularization adds the square of coefficients, leading to smaller coefficients without sparsity.

• L1 regularization is used for classification tasks, whereas L2 is for regression only.

What is the purpose of the softmax function in neural networks?

• The purpose of the softmax function is to convert logits into probabilities for multi-class classification.

• To reduce the dimensionality of the input features.

• To normalize the input data for better training.

• To initialize the weights of the neural network.