RINO
Neural network Reachability Neural network verification
?
- Open-loop or closed loop system, can be discrete or continuous-time
- Optional: Neural network which can be used as some inputs of the closed-loop system
- Optional: Configuration file to set initial values, input, and disturbances ranges, and parameters of the analysis
- System: C++
- Neural network: format inspired by the format used by [[Sherlock]]
- Configuration file: ?
Computes inner and outer approximations of reachable sets.
Inner and outer-approximations of the projection on each component of ranges, and joint 2D and 3D inner-approximations. It also computes approximations of output ranges that are reachable robustly or adversarially w.r.t. disturbances, specified as a subset of inputs.
Uses [[FILIB++]] library for interval computations, [[aaflib]] library for affine arithmetic and [[FADBAD++]] library for automatic differentiation.
Repository: https://github.com/cosynus-lix/RINO
22 September 2022
7 August 2022
RINO: Robust INner and Outer Approximated Reachability of Neural Networks Controlled Systems (CAV 2022)
Other tools that focus on reachability analysis of neural network controlled systems with smooth activation functions: [[Sherlock]], Flow*, NNV, [[ReachNN]], [[ReachNN*]], Verisig, [[JuliaReach]], [[POLAR]].
:: Neural network :: PV1 :: Computes reachable sets for dynamical systems :: Source :: https://doi.org/10.1007/978-3-031-13185-1