OpenReservoirComputing 0.3.0

GPU accelerated implementations of common RC architectures

ORC: Open Reservoir Computing

ORC is the one-stop-shop for performant reservoir computing in jax. Key high-level features include

• Modular design for mixing and matching layers and reservoir drivers (or creating your own!)
• Continuous, discrete, serial, and parallel implementations

Installation

The easiest way to get started with ORC is to install from PyPI:

pip install OpenReservoirComputing

If you're interested in the latest, unreleased version or in contributing, you can install from source. Please see the Contribution guidelines below for more details.

Quick start example

Below is a minimal quick-start example to train your first RC with ORC. It leverages the built-in data library to integrate the Lorenz63 ODE before training and forecasting with ORC.

import jax
# ORC models often perform much better with float64 enabled
jax.config.update("jax_enable_x64", True)
import orc

# integrate the Lorenz system 
U,t = orc.data.lorenz63(tN=100, dt=0.01)

# train-test split
test_perc = 0.2
split_idx = int((1 - test_perc) * U.shape[0])
U_train = U[:split_idx, :]
t_train = t[:split_idx]
U_test = U[split_idx:, :]
t_test = t[split_idx:]

# Initialize and train the ESN
esn = orc.forecaster.ESNForecaster(data_dim=3, res_dim=400)
esn, R = orc.forecaster.train_RCForecaster(esn, U_train)

# Forecast! 
U_pred = esn.forecast(fcast_len=U_test.shape[0], res_state=R[-1]) # feed in the last reservoir state seen in training

To visualize the forecast and compare it to the test data, we can use orc.utils.visualization:

orc.utils.visualization.plot_time_series(
    [U_test, U_pred],
    (t_test - t_test[0]), # start time at 0
    state_var_names=["$u_1$", "$u_2$", "$u_3$"],
    time_series_labels=["True", "Predicted"],
    line_formats=["-", "r--"],
    x_label= r"$t$",
)
plt.show()

jit, vmap, grad...

ORC models are built on top of Equinox, and as a result we strongly recommend the use of Equinox transforms eqx.filter_{jit, vmap, grad} over jax.{jit, vmap, grad}. For more details, please check out the JAX JIT Compatibility example notebook.

Contribution guidelines

First off, thanks for helping out! We appreciate your willingness to contribute! To get started, clone the repo and install the developer dependencies of ORC.

git clone https://github.com/Jan-Williams/OpenReservoirComputing.git

From the root directory of the repository, create an editable install for your given hardware.

CPU:

pip install -e ".[dev]"

GPU:

pip install -e ".[dev, gpu]"

The main branch is protected from direct changes. If you would like to make a change please create a new branch and work on your new feature. After you are satisfied with your changes, please run our testing suite to ensure all is working well. We also expect new tests to be written for all changes if additions are made. The tests can be simply run from the root directory of the repository with

pytest

Followed by a formatting check

ruff check

and a type annotation check

ty check

Finally, submit your changes as a pull request! When you submit the PR, please request reviews from both @dtretiak and @Jan-Williams, we will try to get back to you as soon as possible. When you submit the PR, the above tests will automatically be run on your proposed changes through Github Actions, so it is best to get everything tested first before submitting!