Live Session - Getting Started with Machine Learning in Unity

Here's a screenshot from the recently completed tutorial. Absolutely brilliant tutorial. Really loved the projects...

Completed the Hummingbirds Ml-agents project with small modifications!

My ML agent for the humming bird tutorial in action

Since the LEGO Microgame has just come out, I decided to use it in order to test the ML Agents skills I've just learned throughout the course. I've created a small environment where both the player and the agent have to collect fruit and place it inside a cage, while avoiding crabs, which take away 1 point form the score. Whoever scores the most points when time runs out wins the match. It wasn't easy to make the agent behave as I intended, but after a few tweaks and training sessions, I managed to get a decent neural network result.

Hello, I followed Adam's awesome guide and managed to get the ML Agent working after training it for about 2.5M steps using Adam's settings in trainer_config.yaml. The fact the agent worked despite the fewer steps was a surprise, as Adam trained his for 5M steps(The agent in some runs can get stuck behind or below the red petals, maybe cause of the fewer training steps but works fine most of the time). This course was fun and I wanna thank Unity and Adam Kelly for the free course. My work isn't original, am a beginner and will need more practice to make my own. Either way, it was pretty cool to learn and see the Hummingbird seek and drink nectar on its own. BTW, for anyone following this course recently. The trainer_config.yaml available in the course material won't execute. You will need to update the trainer_config file. The fastest working fix I found online was: https://forum.unity.com/threads/trainerconfigerror-the-option-default-was-specified-in-your-yaml-file-but-is-invalid.912671/ Scroll down to a post by Point_Nemo.

Ball learned not to fall by MLAgents. 5 observation and 1 action