🚀 AI Interview Questions with Answers — Part 9

81. What is Reinforcement Learning?

Reinforcement Learning (RL) is a type of Machine Learning where an agent learns by interacting with an environment and receiving rewards or penalties.

Goal 
Maximize cumulative rewards over time.

Main Components 
Agent → Learner/decision maker 
Environment → Surroundings 
Action → Decision taken 
Reward → Feedback received 

How It Works 
1. Agent takes action
2. Environment responds
3. Agent receives reward or penalty
4. Agent improves strategy

👉 Example: AI learning to play chess through trial and error.

82. What is an agent in Reinforcement Learning?

An agent is the entity that interacts with the environment and makes decisions.

Responsibilities of an Agent 
• Observe environment
• Take actions
• Learn from rewards
• Improve future decisions

Examples 
• Self-driving car
• Robot
• AI game player

👉 Example: In a chess game: 
AI player = Agent 
Chessboard = Environment 

83. What is a reward function?

A reward function defines the feedback an agent receives after taking an action.

Purpose 
Guide the agent toward desired behavior.

Examples 
• Positive reward → Correct action
• Negative reward → Wrong action

Example in Gaming 
Winning a game → +100 reward 
Losing → -100 penalty 

The agent learns strategies that maximize rewards.

84. What is a policy in Reinforcement Learning?

A policy is the strategy an agent follows to decide actions.

It maps: 
States → Actions

Types of Policies 
• Deterministic Policy
• Stochastic Policy

Goal 
Find the optimal policy that gives maximum rewards.

👉 Example: A robot learning the best path to reach a destination.

85. What is the exploration vs exploitation tradeoff?

This tradeoff describes whether the agent should: 
• Explore new actions OR
• Exploit known successful actions

Exploration 
Try new possibilities to gather knowledge.

Exploitation 
Use known best actions for maximum reward.

Challenge 
Balance both effectively.

👉 Example: In gaming: 
Exploring → Trying new moves 
Exploiting → Using proven winning moves 

86. Can you explain Q-Learning?

Q-Learning is a popular Reinforcement Learning algorithm that learns the value of actions in different states.

It uses a Q-table to store values.

Q-Value Formula 
Q(s,a) = Q(s,a) + α[r + γ max Q(s',a') - Q(s,a)] 
Where: 
• Q(s,a) = Current Q-value
• α = Learning rate
• r = Reward
• γ = Discount factor

Goal 
Learn the best action for every state.

👉 Example: AI learning the shortest route in a maze.

87. What is the difference between Reinforcement Learning and supervised learning?

Reinforcement Learning vs Supervised Learning 
Reinforcement Learning - Learns through rewards 
Supervised Learning - Learns from labeled data 

Reinforcement Learning - No correct answers provided directly 
Supervised Learning - Correct answers already available 

Reinforcement Learning - Focuses on sequential decisions 
Supervised Learning - Focuses on predictions 

Reinforcement Learning - Trial-and-error learning 
Supervised Learning - Pattern learning 

Examples 
RL → Game playing AI 
Supervised → Spam detection 

88. What are some real-world applications of Reinforcement Learning?

Applications of RL

1. Self-driving Cars 
Learning safe driving strategies.

2. Robotics 
Robots learning movements and tasks.

3. Gaming 
AI mastering games like chess and Go.

4. Recommendation Systems 
Optimizing user recommendations.

5. Finance 
Automated trading systems.

👉 Example: DeepMind used RL to build AlphaGo, which defeated world champions in Go.

89. What is Deep Q Network (DQN)?

Deep Q Network (DQN) combines: 
• Q-Learning
• Deep Neural Networks

Instead of storing Q-values in tables, it uses neural networks to approximate them.

Advantages 
• Handles large state spaces
• Learns complex patterns
• Better scalability

Applications 
• Gaming AI
• Robotics
• Autonomous systems

👉 Example: AI playing Atari games using Deep Learning.