boundlab

BoundLab - A framework for building neural network verification tools.

Examples

Create a simple zonotope input and concretize bounds:

>>> import torch
>>> import boundlab.expr as expr
>>> x = expr.ConstVal(torch.tensor([0.0, 1.0])) + expr.LpEpsilon([2])
>>> x.ublb()[0].shape
torch.Size([2])

Modules

linearop	Linear Operator Library for Expression Backpropagation.
expr	Symbolic Expression Framework for Bound Propagation
prop	Bound Propagation for Concretizing Expressions
interp	Abstract Interpretation Framework for Neural Network Verification.
zono	Zonotope-Based Abstract Interpretation for Neural Networks
poly	Polytope-Based Abstract Interpretation for Neural Networks.
diff	Toolkits for differential verification.
utils
gradlin	Batched trapezoid linearization for unary functions.