Live Session - Getting Started with Machine Learning in Unity
Exercise
·
Beginner
·
+180XP
·
60 mins
·
Unity Technologies
In this recording of the live session, you'll learn how to create a machine learning experience that has a basic input (one or two dimensional vector observation), an obvious reward, and simple reset state. Throughout the session, learners will be able to:
- Define Machine Learning
- Discuss how ML differs from and works with Artificial Intelligence
- Identify observations for an ML agent to make in a given Scene
- Identify ways to reward the ML agent that will reinforce desired behavior
- Script the ML Agent to collect observations about conditions in the scene and calculate rewards accordingly
- Script the ML Agent to change behavior based on rewards received
Languages available:
1. Live Session Recording - October 20, 2020
Download the screenshot of challenge #4 from the Slides here.
2. Share your project!
We would love to see what you were able to create after learning about Machine Learning! Please take a screenshot of your project or do a screen-recording walking us through it, then post it here to share your creation.
Please login to submit
Submission Gallery
Basic RL project
The agent needs to choose walk left, right or stay still. Getting on any of 2 rewards finish the game Based on Basic project of https://github.com/Unity-Technologies/ml-agents For my buildWebGL: https://play.unity.com/en/games/6b186533-4c65-4386-a13f-685bb36cf519/simple-rl-example GitHub: https://github.com/TitoBotelho/ml-agents-library-unity
Obstacle Dodger
In this game Agent Dodge Obstacles
Kart Agent
SAC vs PPO
ML Drone Agent
It is a drone that uses ray sensors to locate and identify its target, and once it has identified it, it tries to fly as close as possible and ram into the target.
MLAgents Hummingbird Trained Agent
Used MLAgents and Unity to train an agent in the form of a hummingbird to collect nectar from flowers
Galaxy Shooter AI
Not long ago I took a Unity course on Udemy in which the project was a spaceship game. Now I'm applying ML-Agents to create an Artificial Intelligence for the enemies and generate the game's bosses.
Humming bird
my project
CAMERA image as an input ( 36 * 36 pixels )
I improved project for hummingbird to be able to use camera input. I used also immitation learning (when you record your actions and teach NN to act closely as recorded). And even using it last iteration learned for 8 days! (Mac M1). If you have any questions or know how to make it better, write me please www.linkedin.com/in/vshirokun ... Project is here: drive.google.com/file/d/1O_vkL4-LBKStT1a9kRwXIlaLyA67QlSH/view