WS-2025 AI-ML Review Test

What is the primary purpose of a mathematical model in machine learning?

• To replicate physical systems in a laboratory

• To optimize hardware performance

• To represent patterns or relationships within data

• To replace all programming tasks

Which type of function is commonly used in regression models in AI/ML?

• Exponential function

• Sigmoid function

• Linear function

• Step function

In supervised learning, the loss function is used to:

• Randomly initialize model parameters

• Measure the difference between predicted and actual values

• Add noise to the training data

• Generate new datasets for testing

The gradient descent algorithm in machine learning is used to:

• Find the maximum of a loss function

• Optimize model parameters by minimizing the loss function

• Generate predictions for test data

• Normalize input data

Which mathematical concept is central to neural networks?

• Probability theory

• Linear algebra

• Differential equations

• Game theory

What is the primary building block of a deep neural network?

• Gradient descent

• Activation function

• Artificial neuron

• Loss function

Which activation function is most commonly used in hidden layers of deep learning models?

• Sigmoid

• ReLU (Rectified Linear Unit)

• Tanh

• Step function

What is overfitting in deep learning?

• A model that performs poorly on both training and testing data

• A model that generalizes well to unseen data

• A model that performs very well on training data but poorly on testing data

• A model that takes too long to train

In convolutional neural networks (CNNs), what is the primary purpose of the pooling layer?

• To reduce the dimensionality of feature maps

• To learn feature representations

• To add non-linearity to the model

• To initialize weights

What is the key idea behind transfer learning in AI?

• Training a model from scratch for each task

• Using a pre-trained model on a new but related task

• Applying reinforcement learning to optimize performance

• Sharing data between different neural networks