Licensing

The majority of neural-scs is licensed under the CC BY-NC 4.0 License, however, portions of the project are available under separate license terms: SCS is licensed under MIT license.

Neural Fixed-Point Acceleration for SCS

We present neural fixed-point acceleration, a framework to automatically learn to accelerate convex fixed-point problems that are drawn from a distribution, using ideas from meta-learning and classical acceleration algorithms. We apply our framework to SCS, the state-of-the-art solver for convex cone programming. Our work brings neural acceleration into any optimization problem expressible with CVXPY.

Requirements

The following packages are required to run our code:

torch
numpy
scipy
matplotlib
cvxpy
tensorboard
hydra-core
pandas