AI Training and Model Terminology Quiz

1. What does 'batch size' refer to in training?

The number of samples processed before updating model weights

The total number of training samples available

The number of layers in a neural network

The amount of memory required for inference

2. In machine learning, what does 'convergence' indicate?

The training process has stabilized and loss is no longer significantly decreasing

The model has achieved 100% accuracy on training data

All hyperparameters have been optimized

The model has been deployed to production

3. What is the primary goal of cross-validation?

To assess model performance and reduce variance in performance estimates

To increase the size of the training dataset

To automatically tune all hyperparameters

To eliminate overfitting completely

4. Which concept refers to a model's ability to perform well on new, unseen data?

Generalization

Memorization

Convergence

Optimization

5. What does 'hyperparameter' mean in machine learning?

A parameter set before training that controls model behavior

A weight learned during the training process

A metric that measures model performance

A constraint applied during inference

6. In the context of neural networks, what is an 'activation function'?

A non-linear function applied to neuron outputs to enable complex pattern learning

A function that determines which neurons to remove

A method for normalizing input data

A technique for parallelizing computations

7. What does 'gradient descent' accomplish?

Iteratively reduces loss by updating weights in the direction of steepest descent

Measures the accuracy of model predictions

Selects features from raw data

Divides data into training and test sets

8. What is the difference between validation and test data?

Validation tunes hyperparameters; test evaluates final performance on unseen data

They serve the exact same purpose

Test data is used during training; validation is not

Validation data is larger than test data

9. Which of the following best describes 'feature engineering'?

Selecting and transforming raw data into meaningful inputs for model training

Automatically generating all possible model architectures

Reducing the number of model layers

Increasing the learning rate during training

10. What is an 'epoch' in model training?

One complete pass through the entire training dataset

A single update of model weights

A validation checkpoint during training

The final evaluation on test data

11. What is the primary purpose of training in machine learning?

To adjust model weights and parameters using labeled data

To evaluate model accuracy on new unseen data

To deploy the model to production servers

To reduce computational costs during execution

12. In deep learning, what is a 'layer'?

A collection of neurons that process inputs and pass outputs to the next layer

A single neuron in the network

A storage mechanism for training data

A method for splitting data into batches

13. What is the purpose of regularization in machine learning?

To reduce overfitting by penalizing complex models

To speed up the training process

To increase the number of model parameters

To eliminate the need for validation data

14. Which optimization algorithm is commonly used to minimize loss during training?

Stochastic Gradient Descent (SGD)

Principal Component Analysis (PCA)

K-means clustering

Cross-validation

15. What does the term 'loss function' refer to?

A metric that quantifies prediction error during training

The final accuracy score on test data

The computational cost of running inference

The number of parameters in a neural network

16. In supervised learning, what role does labeled data play?

Provides input-output pairs to guide model learning

Helps identify clusters of similar data points

Reduces the need for model validation

Eliminates the need for hyperparameter tuning

17. What is overfitting in machine learning?

When a model learns noise and details specific to training data, reducing generalization

When a model fails to learn any patterns from training data

When a model trains too quickly

When model weights become too large

18. Which term describes the process of using a trained model to make predictions on new data?

Inference

Validation

Regularization

Normalization

19. What is backpropagation?

An algorithm for computing gradients and updating weights during training

A method for storing previous model versions

A technique for increasing training speed

A way to compress model size for inference

20. In the context of neural networks, what does a neuron (or unit) compute?

A weighted sum of inputs followed by a non-linear activation function

Only the raw sum of all input values

A simple counting mechanism for data samples

The final output classification label

SecAI+ Training Inference and Model Terminology Quiz