build details

Show: section status errors & todos local changes recent changes last change in-page changes feedback controls

Imitation Learning

Modified 2020-11-29 by Aleksandar Petrov

Train your network by imitation.

In this part you will try to train a network via imitation, and consequently test its performance.

Some theory about machine learning

A proper laptop setup.

Get a feeling of what imitation learning is.

Learning imitation learning

Modified 2019-10-29 by Gianmarco Bernasconi

Imitation Learning

For this exercise, you will get a repository that takes care of creating the environment for you. You will be required to modify a Jupyter Notebook file only, as the rest is fairly complex.

Clone the template repository:

laptop $ git clone --recursive git@github.com:duckietown-ethz/lra-exercise.git

The --recursive option is important to also clone submodules, if you forgot it, run

laptop $ git submodule init

laptop $ git submodule update

Inside the repository, requirements

laptop $ docker-compose build
laptop $ docker-compose up

The build command, will require a lot of time, but you only need to build it once.

A lot of text will pop up, connect to the notebook server, to do this, you should see an address of the form http://127.0.0.1:8888/..., just open the link in a web browser. Once connected enter in the notebook directory and open the file 01-duckietown-imitation-learning-ipynb, then follow the instructions in it.

When you want to interrupt the containers, press Ctrl-C then wait for it to gracefully stop. This will prevent many problems.