dinos 0.1.1

Active Learning Environments

Dinos

Dinos is a simulation environment for active learning algorithms.

Getting started

First of all, install the package either using pip:

pip install dinos

Or from the git repository:

pip install -r ./requirements.txt
pip install -e .

Examples are provided in the examples folder from the git repository.

How does it works

To run a Dinos experiment you need an Environment and an Agent.
For instance, an environment may be initialized as follow:

from dinos.environments.playground import PlaygroundEnvironment
env = PlaygroundEnvironment()

From there you can either use your own code and use low level API to interact with the environment: env.step(self, action, actionParameters=[], config=None) as detailed later on. The second option is to use the Dinos Agent system to manage your algorithm.

For instance to create an agent that will perform a random action at each step:

from dinos.agents.random import RandomAgent
agent = RandomAgent(env.world.findHost())

env.world.findHost() let you find an entity in the environment that can be controlled by your learner (we call such entity an host)

Each Agent has a reach(self, configOrGoal) method that can be used to tell the agent to reach a specific goal.

Additionally a specific type of agent exists: Learner. This class is designed to be used with a dataset or a memory to learn from its interactions with the environment.

Each Learner has a train(self, iterations=None, untilIteration=None, episodes=None, untilEpisode=None) method used to train your learner for a given number of iterations or episodes.

More details are present in the examples folder from the git repository.