Examples

We provide a number of reference examples directly on our GitHub. Each of these examples demonstrates how to recreate a particular experiment or result from recent evolutionary algorithm literature, to highlight that Evotorch is highly suited to both academic research in and advanced industrial applications of evolutionary algorithms.

Tip

We provide a number of examples as jupyter notebooks. The easiest way to get started with these examples is to run:

cd evotorch
pip install jupyterlab
jupyter lab

Examples in examples/notebooks

Gym Experiments with PGPE and CoSyNE

Demonstrates how you can solve "LunarLanderContinuous-v2" using both PGPE and CoSyNE following the configurations described in the paper proposing ClipUp and the JMLR paper on the CoSyNE algorithm.

Minimizing Lennard-Jones Atom Cluster Potentials

Recreates experiments from the paper introducing SNES, showing that the algorithm can effectively solve the challenging task of minimising Lennard-Jones atom cluster potentials.

Model_Predictive_Control_with_CEM

Demonstrates the application of the Cross-Entropy Method CEM to Model Predictive Control (MPC) of the MuJoCo task named "Reacher-v4".

Training MNIST30K

Recreates experiments from a recent paper which demonstrates that SNES can be used to solve supervised learning problems. The script in particular recreates the training of the 30K-parameter 'MNIST30K' model on the MNIST dataset, but can easily be reconfigured to recreate other experiments from that paper.

Variational Quantum Eigensolvers with SNES

Re-implements (with some minor changes in experimental setup), experiments in a recent paper demonstrating that SNES is a scalable alternative to analytic gradients on a quantum computer, and can practically optimize Quantum Eigensolvers.

Examples in examples/scripts

In addition, to help you to implement advanced neuroevolutionary reinforcement learning settings, we have provided 3 python scripts in the examples/scripts directory:

bbo_vectorized.py

Demonstrates single objective black-box optimization using a distribution-based algorithm, accelerated using vectorization on a single GPU/CPU.

moo_parallel.py

Demonstrates multi-objective optimization using parallelization on all CPU cores without vectorization.

rl_clipup.py

Re-implements almost all experiments from the paper proposing ClipUp, and is easily reconfigured to replicate any particular experiment using sacred.

rl_enjoy.py

Allows you to easily visualize and enjoy agents trained through rl_clipup.py.

rl_gym.py

Demonstrates how to solve a simple Gym problem using the PGPE algorithm and ClipUp optimizer.