evox
Project description
EvoX is a distributed GPU-accelerated framework for scalable evolutionary computation. Our primary goal is to push the boundaries of evolutionary computation by significantly enhancing its speed and versatility, enabling its application to complex and computationally intensive tasks.
⭐️ Key Features
- 🚀 Fast
- GPU computing for 10x-100x faster optimization.
- Distributed workflow for even faster optimization.
- 🌟 Wide support
- Single-objective and multi-objective optimization.
- Comprehensive support for commonly used benchmark problems.
- Extensive coverage of neuroevolution problems.
- 🎉 Easy to use
- Functional programming for easy function composition.
- Hierarchical state management for modular programming.
- Detailed tutorial available here.
EvoX offers a powerful and user-friendly optimization framework, empowering researchers and practitioners to easily tackle a variety of optimization tasks. The support for commonly used benchmark problems, along with the coverage of neuroevolution problems, provides a versatile platform for optimization experimentation. With its fast GPU computing and distributed workflow capabilities, EvoX enables efficient optimization of complex and computationally intensive problems. The functional programming and hierarchical state management further enhance the ease of use and modularity of the framework.
Index
Contents
List of Algorithms
Single-objective
Type | Algorithm Name |
---|---|
Differential Evolution | CoDE, JaDE, SaDE, SHADE, IMODE, ... |
Evolution Strategies | CMA-ES, PGPE, OpenES, CR-FM-NES, xNES, ... |
Particle Swarm | FIPS, CSO, CPSO, CLPSO, SL-PSO, ... |
Multi-objective
Type | Algorithm Name |
---|---|
Dominance-based | NSGA-II, NSGA-III, SPEA2, BiGE, KnEA, ... |
Decomposition-based | MOEA/D, RVEA, t-DEA, MOEAD-M2M, EAG-MOEAD, ... |
Indicator-based | IBEA, HypE, SRA, MaOEA-IGD, AR-MOEA, ... |
List of Problems
Type | Problem Name |
---|---|
Numerical | DTLZ, LSMOP, MaF, ZDT, CEC'22, ... |
Neuroevolution | Brax, Gym, TorchVision Dataset, ... |
For more detailed list, please refer to our API documentation. List of Algorithms and List of Problems.
Installation
We recommand install evox
using pip
pip install evox
EvoX depends on JAX. To install JAX, please refer to JAX's installation guide here.
Quick Start
To start with, import evox
import evox
from evox import algorithms, problems, workflows
Then, create an algorithm and a problem:
pso = algorithms.PSO(
lb=jnp.full(shape=(2,), fill_value=-32),
ub=jnp.full(shape=(2,), fill_value=32),
pop_size=100,
)
ackley = problems.numerical.Ackley()
To run the EC workflow, compose the algorithm and the problem together using workflow
:
workflow = workflows.StdWorkflow(pso, ackley)
To initialize the whole workflow, call init
on the workflow object with a PRNGKey. Calling init
will recursively initialize a tree of objects, meaning the algorithm pso and problem ackley are automatically initialize as well.
key = jax.random.PRNGKey(42)
state = workflow.init(key)
Now, call step
to execute one iteration of the workflow.
# run the workflow for 100 steps
for i in range(100):
state = workflow.step(state)
Example
The example folder has many examples on how to use EvoX.
Support
- For general discussion, please head to Github's discussion
- For Chinese speakers, please consider to join the QQ group to discuss. (Group number: 297969717).
Citation
@article{evox,
title = {{EvoX}: {A} {Distributed} {GPU}-accelerated {Framework} for {Scalable} {Evolutionary} {Computation}},
author = {Huang, Beichen and Cheng, Ran and Li, Zhuozhao and Jin, Yaochu and Tan, Kay Chen},
journal = {arXiv preprint arXiv:2301.12457},
eprint = {2301.12457},
year = {2023}
}
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.