This is easily the best resource on q-learning I’ve ever seen. It’s one simple library available to everyone on any OS (looking at you, *gym* ) and very well explained
WOW!!! Thanks a ton for this video mate. I have taken the course in reinforcement learning at university and this is by far the great way to make acquainted with Q-learning algorithm in reinforcement learning.
I'm really impressed by how easily you explained this. I'm going to watch the other videos in your Q-learning series, and subscribe as well. Thanks for this amazing RL tutorial, man.
Thankyou so much for this clear and practical example. Most videos I've seen just kind of breeze over the environment and states. Keep up the good work!
Thank you very much Dr. Soper! I took great value from this video, as I am trying to implement some Q-Learning techniques into a project of my own. Looking forward to your future videos!
Very easy to follow and understand Q-learning and see amazing computer 💻 working so well. Encourage myself to learning more about AI. I am not good at maths, and logical problems but I understand easily by your videos. Thank you so much 😊
Very helpful and engaging lesson. Was having hard time understanding implementation but this visual explanation with code implementation made it very clear. Great job!!
it is very good that this can automatically show us the shortest path but what if we need to know the q table or the updated table which program uses to find the best action from up down right or left. i meant any action is taken from the updated table. if we can obtain that it is a huge success as well.
Gemini: This video is about a complete walkthrough of a Q-learning based AI system in Python. The video starts with an introduction to the business problem. The problem is about designing a warehouse robot that can travel around the warehouse to pick up items and bring them to a packaging area. The robot needs to learn the shortest path between all the locations in the warehouse. Then the video explains the concept of Q-learning, which is a reinforcement learning technique. Q-learning works by letting an agent learn from trial and error. The agent receives rewards for taking good actions and penalties for taking bad actions. Over time, the agent learns to take the actions that will lead to the greatest reward. Next, the video dives into the code. The code defines the environment, which includes the states, actions, and rewards. The states are all the possible locations of the robot in the warehouse. The actions are the four directions that the robot can move (up, down, left, and right). The rewards are positive for reaching the packaging area and negative for all other locations. The code also defines a Q-learning agent. The agent starts at a random location in the warehouse and then takes a series of actions. The agent learns from the rewards that it receives for its actions. Over time, the agent learns to take the shortest path to the packaging area. Once the agent is trained, the video shows how to use the agent to find the shortest path between any two locations in the warehouse. The video also shows how to reverse the path so that the robot can travel from the packaging area to any other location in the warehouse. Overall, this video is a great introduction to Q-learning and how it can be used to solve real-world problems.
My humble thanks Dr.Daniel for such a clear description of Q-learning in python. I am not able to access the notebook for the code. Kindly could you help me regarding this. I want to practice programming for creating an environment and execute q-leaning on it.
Have you installed all the required libraries and are running the correct version of python? And if your running the code on the website run the code in order. Edit: try running it natively on your computer
Thank you for the video. It was really clear and helpful. I have one question. In obtaining the shortest path after training, could you explain why epsilon is 0.9? Shouldn't the epsilon = 1.0 to maximize the rewards?
He explains it briefly in the video at 13:00. The way I understand it is that you do sometimes want the AI agent to take random actions, instead of always taking the best action, so that it has a chance to explore the environment. Otherwise the AI, as soon as it finds a route to the 'item packaging area', might think it has already found the best route, and stop looking for a better one.
Thank you Dr. Daniel ! This is a excellent Q-leaning instructional video includ comprehensive theory part and practical implementation. I want to inquiry is it still possible to find the link to this notebook now ? I don't find where is the video descroption part.
Thanks for video! However, I don't think that making every navigable tile have a -1 is optimal. If the robot has a bitmap of the environment, and can reference the location it is at, then can base a navigable tiles reward system that increments up to MUCH better GUIDE the robot toward the goal. Rather than a binary goal.
This is easily the best resource on q-learning I’ve ever seen. It’s one simple library available to everyone on any OS (looking at you, *gym* ) and very well explained
Wow this is the best qlearning video with code example that i found on the internet. Thanks for the detailed explanation of every step in the code
I don't believe anyone teaches it better than you. Amazing.
Best Q-Learning lesson ever, better than my uni class. Thank you very much, please keep it up.
I am learning RL for last 1 year and this is one of the best video. You have taught everything from start to end. Thank you!
Extremely clear explanation for this topic. You are my life saver when I am preparing my finals. Thanks a lot.
The code is much cleaner and easier to understand than the example codes I found in many other Q-Learning books.
That is some incredible teaching skill! Broke down a complex concept just to the right amount of detail . Really appreciate this lesson.
Thank you so much Dr.Super! You're a very gifted teacher. Please don't stop.
Best RL video ever in UA-cam! Thank you so much, Dr. Soper!
WOW!!!
Thanks a ton for this video mate.
I have taken the course in reinforcement learning at university and this is by far the great way to make acquainted with Q-learning algorithm in reinforcement learning.
The best teacher i never seen. Respect.
You are an excellent teacher, Dr. Soper. Appreciate your support and presentation.
I'm really impressed by how easily you explained this. I'm going to watch the other videos in your Q-learning series, and subscribe as well. Thanks for this amazing RL tutorial, man.
i never expected to see fl studio tutorial channel here lol
Thankyou so much for this clear and practical example. Most videos I've seen just kind of breeze over the environment and states. Keep up the good work!
Dude thanks a bunch. You gave just the right amount of detail and broke it down simply. Thanks for not bogging us down with a ton of details.
Writing my dissertation and this is a God-send. Thank you :)
Thank you for this video, you've explained this excellently and I actually understand the concept now!
Thank you very much Dr. Soper! I took great value from this video, as I am trying to implement some Q-Learning techniques into a project of my own. Looking forward to your future videos!
i am really enjoying going through your videos.
Very easy to follow and understand Q-learning and see amazing computer 💻 working so well. Encourage myself to learning more about AI. I am not good at maths, and logical problems but I understand easily by your videos. Thank you so much 😊
Very helpful and engaging lesson. Was having hard time understanding implementation but this visual explanation with code implementation made it very clear. Great job!!
Excellent explanation with such a pleasant voice! Thank you so much.
Such a complex topic is explained without any hassle!
Awesome! Thank you very much, so intuitive and easy to understand video!!
You sir THANK YOU.. you broke it down very easy to comprehend and learn. Thank you and thank you
Best Q - Learning lesson, congratulation and thank u
So peaceful yet very informative. Love this style.
Amazing video. Better than my uni explanation!
Perfect tutorial with clear sample code, but only 646 views? This video deserves better!
Most other videos are just using GYM or Unity library for their video, which are just show-offs, instead of really teaching something.
very beneficial for my master's thesis. THANKS!!
Thanks man this video helped out by quite a lot. keep up the good work
Fantastic lesson!! Thanks so much
Excellent lesson!Thank you so much
error in the get_next_location function. actions not define
Amazing explanation. Thanks a lot.
Great video, very well made.
why the agent gets -1 in its walking area.
it is very good that this can automatically show us the shortest path but what if we need to know the q table or the updated table which program uses to find the best action from up down right or left. i meant any action is taken from the updated table. if we can obtain that it is a huge success as well.
when it comes to a certain point it has to decide where to go so updated table shows the values of each direction, highest value will be selected
WHAT AN AWESOME VIDEO
Your are amazing sir, keep up the good work, thank you
what type of q learning is used in this?
Excellent video sir
Gemini: This video is about a complete walkthrough of a Q-learning based AI system in Python.
The video starts with an introduction to the business problem. The problem is about designing a warehouse robot that can travel around the warehouse to pick up items and bring them to a packaging area. The robot needs to learn the shortest path between all the locations in the warehouse.
Then the video explains the concept of Q-learning, which is a reinforcement learning technique. Q-learning works by letting an agent learn from trial and error. The agent receives rewards for taking good actions and penalties for taking bad actions. Over time, the agent learns to take the actions that will lead to the greatest reward.
Next, the video dives into the code. The code defines the environment, which includes the states, actions, and rewards. The states are all the possible locations of the robot in the warehouse. The actions are the four directions that the robot can move (up, down, left, and right). The rewards are positive for reaching the packaging area and negative for all other locations.
The code also defines a Q-learning agent. The agent starts at a random location in the warehouse and then takes a series of actions. The agent learns from the rewards that it receives for its actions. Over time, the agent learns to take the shortest path to the packaging area.
Once the agent is trained, the video shows how to use the agent to find the shortest path between any two locations in the warehouse. The video also shows how to reverse the path so that the robot can travel from the packaging area to any other location in the warehouse.
Overall, this video is a great introduction to Q-learning and how it can be used to solve real-world problems.
very well explained.. Thankyou
Thanks for easy presentation
can u do deep q learning complete example please? i really need that one :)
So helpful! Thank you!
this is awesome. I wonder how long it to took to do all this.
the background music for this is just perfect! (Anybody knows if it is something easily accesable?)
My humble thanks Dr.Daniel for such a clear description of Q-learning in python.
I am not able to access the notebook for the code.
Kindly could you help me regarding this. I want to practice programming for creating an environment and execute q-leaning on it.
AMAZING!!!! Thanks a lot
Appreciate your knowledge sharing.
It is not working for me. (code on the website has errers)
Have you installed all the required libraries and are running the correct version of python? And if your running the code on the website run the code in order.
Edit: try running it natively on your computer
This is quite strange to me, it seems that the agent does not use any of the q-value to create the shortest path
Really good video and helped a lot!
However the piano is a bit too loud :D maybe -20%? :D
Great content
Thank you for the video. It was really clear and helpful. I have one question.
In obtaining the shortest path after training, could you explain why epsilon is 0.9? Shouldn't the epsilon = 1.0 to maximize the rewards?
He explains it briefly in the video at 13:00. The way I understand it is that you do sometimes want the AI agent to take random actions, instead of always taking the best action, so that it has a chance to explore the environment. Otherwise the AI, as soon as it finds a route to the 'item packaging area', might think it has already found the best route, and stop looking for a better one.
Thank you Dr. Daniel ! This is a excellent Q-leaning instructional video includ comprehensive theory part and practical implementation. I want to inquiry is it still possible to find the link to this notebook now ? I don't find where is the video descroption part.
I find where the link is, thx :)
this is great!!!!!!!!
Thanks for video! However, I don't think that making every navigable tile have a -1 is optimal. If the robot has a bitmap of the environment, and can reference the location it is at, then can base a navigable tiles reward system that increments up to MUCH better GUIDE the robot toward the goal. Rather than a binary goal.
Get rid of the background noise
What a legend
Thank you
4:50 poor robots, they are never going to be good enough :(
background music in very annoying! But lesson was good
Really nice lesson , my unversity prof is so bad compared to this.