Challenging ML Concepts for Students

What are the key differences between feedforward and recurrent neural networks?

• Feedforward networks are used for time series prediction.

• Feedforward networks have no cycles and process data in one direction, while recurrent networks have loops that allow them to maintain memory of previous inputs.

• Recurrent networks do not have memory of previous inputs.

• Feedforward networks can process data in multiple directions.

Explain the concept of the kernel trick in Support Vector Machines.

• The kernel trick allows SVM to classify data in higher-dimensional spaces using kernel functions, enabling non-linear separation without explicit transformation.

• The kernel trick simplifies SVM by reducing the number of dimensions needed for classification.

• The kernel trick is a method to visualize data in two dimensions only.

• The kernel trick is used to transform data into a linear format for easier classification.

What is the purpose of pruning in decision trees, and how does it affect model performance?

• Pruning has no impact on the model's accuracy.

• Pruning increases the complexity of the decision tree.

• The purpose of pruning in decision trees is to reduce overfitting and improve model generalization.

• Pruning is used to enhance the training data size.

Describe the exploration-exploitation trade-off in reinforcement learning.

• The trade-off is irrelevant in reinforcement learning as all actions yield the same rewards.

• Exploration involves only using known actions, while exploitation is about trying new actions.

• The exploration-exploitation trade-off is about maximizing rewards without any risk.

• The exploration-exploitation trade-off is the balance between trying new actions to discover their rewards (exploration) and using known actions that yield high rewards (exploitation) in reinforcement learning.

What are convolutional neural networks primarily used for, and how do they differ from traditional neural networks?

• Convolutional Neural Networks are primarily used for image and video recognition, and they differ from traditional neural networks by using convolutional layers to detect features and reduce the number of parameters.

• Convolutional Neural Networks are primarily used for financial forecasting and stock market analysis.

• They differ from traditional neural networks by using recurrent layers instead of convolutional layers.

• Convolutional Neural Networks are mainly used for natural language processing tasks.

How does the margin in Support Vector Machines influence the decision boundary?

• The margin only influences the training time of the model.

• A larger margin always leads to overfitting.

• The margin has no effect on the decision boundary.

• The margin influences the decision boundary by determining its distance from the nearest data points, affecting the model's generalization ability.

What is the role of ensemble methods like Random Forests in improving model accuracy?

• Ensemble methods like Random Forests only use a single decision tree for predictions.

• Ensemble methods like Random Forests improve model accuracy by combining multiple decision trees to reduce overfitting and variance.

• Random Forests improve accuracy by increasing the depth of individual trees.

• Ensemble methods like Random Forests eliminate the need for data preprocessing.

Define the term 'state' in the context of reinforcement learning and provide an example.

• A state refers to the total number of actions taken; for example, the number of moves in a game.

• A state is the final outcome of an episode; for example, winning a game.

• A state in reinforcement learning is a specific situation in the environment; for example, the arrangement of pieces on a chessboard.

• A state is a fixed rule that governs the environment; for example, the rules of chess.

What is backpropagation, and why is it essential for training neural networks?

• Backpropagation is a method for visualizing neural networks.

• Backpropagation is used to increase the number of neurons in a network.

• Backpropagation is a technique for generating random weights in neural networks.

• Backpropagation is an algorithm for training neural networks by calculating gradients for weight updates.

Explain the difference between supervised and unsupervised learning in the context of decision trees.

• Supervised learning requires no data, while unsupervised learning requires labeled data.

• Supervised learning is only applicable to linear models, while unsupervised learning is for decision trees.

• Both supervised and unsupervised learning use the same type of data for training decision trees.

• Supervised learning uses labeled data for training decision trees, while unsupervised learning uses unlabeled data to find patterns.