Training data and bias flashcards

What is training data?

Data used to teach a machine learning model how to predict outcomes.

True or false: Bias in data is always negative.

False, because bias can sometimes lead to useful simplifications.

Difference between supervised and unsupervised learning.

Supervised learning uses labeled data; unsupervised learning does not.

What is model bias?

A model's tendency to consistently deviate from true values due to training data.

Fill in the blank: __________ can cause bias in AI models.

Imbalanced training data

What is overfitting?

When a model learns noise in training data and performs poorly on new data.

True or false: All datasets are perfectly unbiased.

False, because most datasets contain some form of bias.

Question: How can bias affect AI outcomes?

Bias can lead to unfair or inaccurate predictions, affecting real-world applications.

What is underfitting?

When a model is too simple to capture the underlying patterns in data.

Comparison: Bias vs. Variance.

Bias is error due to oversimplification; variance is error due to complexity.

What is data augmentation?

A technique to increase training data by creating modified versions of existing data.

True or false: Larger datasets always eliminate bias.

False, because larger datasets can still reflect existing biases.

Fill in the blank: __________ is used to correct bias in AI.

Fairness algorithms

Question: What is the role of validation data?

To check how well a model generalizes to new, unseen data.

What does feature selection involve?

Choosing the most relevant attributes from data that contribute to predictions.

True or false: Training data must always be diverse.

True, because diversity helps reduce bias and improve model performance.

What is a data pipeline?

A set of processes that collects, cleans, and prepares data for analysis.