machine_learning
My little personal lab to learn machine learning: I started with a simple Q-learning table, wrote a neural network and a Deep Q-Network in pure Python with no library, and ended with an agent that learns to drive a race car around a track, using virtual distance sensors and laps I drove myself.
Team
Solo
Context
Personal project
Status
Done
Context
A personal project to understand reinforcement learning by building it step by step, each stage reusing what the previous one taught me.
What was built
- Tabular Q-learning agent, to get the basics of rewards, exploration and value updates
- A neural network and a Deep Q-Network written in pure Python, with no library: forward pass, backpropagation and experience replay by hand
- A driving agent in TensorFlow/Keras: a 2D race car with ray sensors, trained with deep Q-learning and kick-started with laps I drove myself