Formal syntax
Defines satisfiability queries in a shared, machine-readable format.
The International Standard for the Verification of Neural Networks
(vnnlib-version <2.0>)
; Network declaration
(declare-network f
(declare-input X float32 [2,2])
(declare-output Y float32 [1])
)
; Input constraints
(assert (and (>= X[0,0] 0.0) (<= X[0,0] 1.0)))
(assert (and (>= X[0,1] 0.0) (<= X[0,1] 1.0)))
(assert (and (>= X[1,0] 0.0) (<= X[1,0] 1.0)))
(assert (and (>= X[1,1] 0.0) (<= X[1,1] 1.0)))
; Output constraints
(assert (or (<= Y[0] -3.0) (>= Y[0] 0.0)))
Introduction
VNN-LIB is an open-source standard for automated solvers of satisfiability problems over neural networks. The goal of the standard is to foster interoperability in the neural network verification community.
Contributions and suggestions to the standard are welcome. If interested, please open an issue describing your proposal on the VNN-LIB Standard GitHub repository.
Defines satisfiability queries in a shared, machine-readable format.
Gives precise meaning to each query so tools interpret problems consistently.
Specifies a command-line interface for exchanging queries with solvers.
Latest News
Team Members
Lecturer | University of Western Australia
PhD Student | University of Genoa
Full Professor | University of Genoa
Masters Student | University of Western Australia
Masters Student | University of Western Australia
Postdoctoral Researcher | University of Genoa
Full Professor | University of Sassari
Postdoctoral Researcher | University of Sassari