Learn Before
Why reward design is hard in reinforcement learning
Question: In one to three sentences, explain why it is hard to write a good reward function for a task such as controlling an autonomous warehouse robot.
Sample answer: A reward function is hard to design because it has to capture several goals at once, such as avoiding collisions, reaching the target quickly, and handling packages carefully. These goals can conflict, so a manually created score can be hard to balance well.
Key points:
- The reward is usually specified by the designer
- It must combine several competing goals, such as safety, speed, and careful handling
- One score may not reflect all of the trade-offs well
Rubric: Full credit: mentions that the reward must balance multiple competing goals and is manually designed. Partial credit: mentions only that it is hand-crafted or only one trade-off issue.
0
1
Tags
Data Science
Foundations of Large Language Models Course
Computing Sciences
Machine Learning
Deep Learning
Supervised Learning
Dive into Deep Learning @ D2L
Machine Learning Yearning @ DeepLearning.AI
Related
Reinforcement Learning Example - Autonomous Vehicles
Reinforcement Learning Analogy - Video Games
Machine Learning for Absolute Beginners
Deep Learning vs. Reinforcement Learning
Fundamental Concepts for Reinforcement Learning
Reinforcement Learning Refrence and Cutting-edge Ideas
A team is developing a program to play a complex board game against human opponents. The program has no pre-existing data of past games to learn from. Instead, it is designed to learn by playing against itself repeatedly. After each game, the program receives a positive signal if it wins and a negative signal if it loses. Over time, it is expected to discover winning strategies on its own. Which of the following statements best analyzes why this learning approach is suitable for this task?
Robot Maze Navigation Strategy
Evaluating Learning Strategies for a Recommendation System
Reward Scores for Reinforcement Learning Trajectories
In reinforcement learning, what is the agent mainly trying to accomplish while it interacts with the environment?
A reinforcement-learning system can keep adapting after deployment if it continues collecting new experience.
In reinforcement learning, an agent tries to maximize the total _____ it earns over time.
Match each reinforcement learning term to its correct description.
Order the steps for training a warehouse robot with reinforcement learning.
Why can reinforcement learning be harder to set up than supervised learning?
Diagnose a flawed reward design for a warehouse drone docking task.
Why reward design is hard in reinforcement learning
Why might a crashed delivery drone receive a large negative reward such as R(T) = -1,000?
True or False: A reward function for teaching a drone to land on a charging pad is usually discovered automatically with no human design.