Implementation of Piecewise Linear Functions (PWL) in PyTorch.
Project description
Piecewise Linear Functions (PWLs) can be used to approximate any 1D function. PWLs are built with a configurable number of line segments - the more segments the more accurate the approximation. This package implements PWLs in PyTorch and as such they can be fit to the data using standard gradient descent. For example:
import torchpwl
# Create a PWL consisting of 3 segments for 5 features - each feature will have its own PWL function. pwl = torchpwl.PWL(num_features=5, num_breakpoints=3) x = torch.Tensor(11, 5).normal_() y = pwl(x)
Monotonicity is also supported via MonoPWL. See the class documentations for more details.
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.