https://github.com/kaustubholpadkar/nevolve
Neuro-Evolution Library for Reinforcement Learning
https://github.com/kaustubholpadkar/nevolve
artificial-intelligence deep-learning evolutionary-algorithms genetic-algorithms machine-learning neural-networks neuroevolution reinforcement-learning
Last synced: 2 months ago
JSON representation
Neuro-Evolution Library for Reinforcement Learning
- Host: GitHub
- URL: https://github.com/kaustubholpadkar/nevolve
- Owner: kaustubholpadkar
- License: gpl-3.0
- Created: 2019-06-16T10:27:47.000Z (over 6 years ago)
- Default Branch: master
- Last Pushed: 2019-09-02T00:50:11.000Z (about 6 years ago)
- Last Synced: 2025-07-02T17:57:33.226Z (3 months ago)
- Topics: artificial-intelligence, deep-learning, evolutionary-algorithms, genetic-algorithms, machine-learning, neural-networks, neuroevolution, reinforcement-learning
- Language: Python
- Size: 333 KB
- Stars: 0
- Watchers: 2
- Forks: 0
- Open Issues: 1
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Nevolve - Neuro-Evolution for Humans
## You have just found Nevolve.
Nevolve is a Neuro-Evolution Library for __Reinforcement Learning__, written in Python, developed aiming at enabling rapid experimentation.
It implements *conventional neuro-evolution algorithm*, which evolves the strength of the connection weights for a fixed network topology. It initializes the Artificial Neural Network with random weights and evolves the weights using Neuro-Evolution.
Read the documentation at [Nevolve.ml](https://nevolve.ml).
Nevolve is compatible with: __Python 3.5 or greater__
## Getting started: 15 seconds to Nevolve
The core class of Nevolve is a __Population__, which organizes whole Neuro-Evolution process.
Here is the `Population` class:
```python
from nevolve.evolve.population import Population
```
A few sample environments are also provided in the library. CartPole environment is one of them. Here is the wrapper on `CartPole` environment from [OpenAi Gym](https://gym.openai.com/):
```python
from nevolve.envs.cart_pole_v0 import CartPoleV0
```
Create Population instance.
```python
model = Population(
cls=CartPoleV0,
size=100,
mutation_rate=0.05,
max_workers=10
)
```
Start Evolution:
```python
model.evolve(
max_generation=10,
show_best=True
)
```
Results:
```
Generation: 1 | Best Fitness: 27.0
Generation: 2 | Best Fitness: 21.0
Generation: 3 | Best Fitness: 41.0
Generation: 4 | Best Fitness: 49.0
Generation: 5 | Best Fitness: 62.0
Generation: 6 | Best Fitness: 48.0
Generation: 7 | Best Fitness: 200.0
Generation: 8 | Best Fitness: 200.0
Generation: 9 | Best Fitness: 200.0
Generation: 10 | Best Fitness: 200.0
```
Once your model performs the best for you, save the best model with `.save_best_()`:
```python
model.save_best_(path="best_cart_pole.pkl")
```
For a more in-depth tutorial about Nevolve, you can check out:
- [Applying Neuro-Evolution on Sample Environments - CartPole, MountainCar & BipedalWalkerV2](https://nevolve.ml/tutorial/example1.html)
- [Create your own Environment wrappers for Neuro-Evolution - Pong Game.](https://nevolve.ml/tutorial/example2.html)
- [Saving and Loading Populations.](https://nevolve.ml/tutorial/example3.html)
------------------
## Installation
- **Install Nevolve from PyPI (recommended):**
Note: These installation steps assume that you are on a Linux or Mac environment.
If you are on Windows, you will need to remove `sudo` to run the commands below.
```sh
sudo pip install nevolve
```
If you are using a virtualenv, you may want to avoid using sudo:
```sh
pip install nevolve
```
- **Alternatively: install Nevolve from the GitHub source:**
First, clone Nevolve using `git`:
```sh
git clone https://github.com/kaustubholpadkar/nevolve.git
```
Then, `cd` to the nevolve folder and run the install command:
```sh
cd nevolve
sudo python setup.py install
```
- **Install OpenAI Gym (optional):**
If you want to run the sample environments or create a wrapper on the environment from OpenAI Gym, you need to install its package using following command.
```sh
pip install gym
```
For more information on OpenAi Gym and installation, please visit [this](http://gym.openai.com/docs/#installation).
------------------
## Future enhancement
- To add support for NEAT algorithm and its variations which can evolve the weights and the topology of the Neural Network.
- To Introduce some computational optimizations in Neural Network library and Evolutionary Algorithm.
- To Introduce a few new wrappers for OpenAI Gym and other environments.
------------------
## Support
You can ask questions and join the development discussion:
- On the [Nevolve Google group](https://groups.google.com/forum/#!forum/nevolve).
- On the [Nevolve Slack channel](https://nevolve.slack.com). Use [this link](https://nevolve-slack-invitation.herokuapp.com/) to request an invitation to the channel.
You can also post **bug reports and feature requests** (only) in [GitHub issues](https://github.com/kaustubholpadkar/nevolve/issues).
------------------
## License
Nevolve is GNU licensed, as found in the LICENSE file.
------------------