*1. What is Artificial Intelligence?*  

Artificial Intelligence (AI) is the branch of computer science that focuses on building machines or systems that can perform tasks that typically require human intelligence — such as understanding language, recognizing images, making decisions, and learning from data.

*2. Difference between AI, Machine Learning, and Deep Learning*  

- *AI*: The broad concept of machines simulating human intelligence.  

- *Machine Learning (ML)*: A subset of AI that enables systems to learn from data and improve over time without being explicitly programmed.  

- *Deep Learning (DL)*: A subfield of ML that uses neural networks with many layers to model complex patterns, especially in images, audio, and text.

*3. What is supervised vs. unsupervised learning?*  

- *Supervised Learning*: The model learns from labeled data. It is trained on input-output pairs.  

  *Example*: Predicting house prices from past data.  

- *Unsupervised Learning*: The model finds patterns in data without labels.  

  *Example*: Grouping customers based on buying behavior (clustering).

*4. Explain overfitting and underfitting*

- *Overfitting*: The model learns noise and details in the training data, performing poorly on new data.  

- *Underfitting*: The model is too simple to capture the data patterns and performs poorly on both training and testing data.  

*A good model generalizes well to unseen data.*

*5. What are classification and regression?*  

- *Classification*: Predicts discrete labels.  

  *Example*: Email spam detection (spam or not).  

- *Regression*: Predicts continuous values.  

  *Example*: Predicting stock price or temperature.

*6. What is a confusion matrix?*  

It’s a table used to evaluate the performance of a classification model by comparing predicted vs. actual results.  

It shows:  

- True Positives (TP)  

- True Negatives (TN)  

- False Positives (FP)  

- False Negatives (FN)

*7. Define precision, recall, F1-score*  

- *Precision* = TP / (TP + FP): How many predicted positives are correct.  

- *Recall* = TP / (TP + FN): How many actual positives are captured.  

- *F1-Score* = Harmonic mean of precision and recall.  

Useful when dealing with imbalanced datasets.

*8. What is the difference between batch and online learning?*  

- *Batch Learning*: The model is trained on the entire dataset at once.

- *Online Learning*: The model is updated incrementally as new data arrives — useful for real-time systems.

*9. Explain bias-variance tradeoff*  

- *Bias*: Error from incorrect assumptions (underfitting).  

- *Variance*: Error from model sensitivity to training data (overfitting).  

*Goal:* Find a balance to minimize total error.

*10. What are activation functions in neural networks?*  

Activation functions decide whether a neuron should fire. They introduce non-linearity into the network.  

Common ones:  

- ReLU: `max(0, x)`  

- Sigmoid: squashes values between 0 and 1  

- Tanh: squashes between -1 and 1