Filters








876 Hits in 9.9 sec

On The Verification of Neural ODEs with Stochastic Guarantees [article]

Sophie Gruenbacher, Ramin Hasani, Mathias Lechner, Jacek Cyranka, Scott A. Smolka, Radu Grosu
2020 arXiv   pre-print
time-horizon), and provide stochastic guarantees in the form of confidence intervals for the Reachtube bounds.  ...  We show that Neural ODEs, an emerging class of time-continuous neural networks, can be verified by solving a set of global-optimization problems.  ...  We use the following to overestimate of Pr(C(r bound )): Using the results of (Shengqiao 2011) it holds that Area(C(r bound )) ≥ Vol n−1 (ρ(r bound )), thus:  ... 
arXiv:2012.08863v1 fatcat:ufxpdbsmfzfnhohcxlni2uguru

On the Verification of Neural ODEs with Stochastic Guarantees

Sophie Grunbacher, Ramin Hasani, Mathias Lechner, Jacek Cyranka, Scott A. Smolka, Radu Grosu
2021 PROCEEDINGS OF THE THIRTIETH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE AND THE TWENTY-EIGHTH INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE  
time-horizon), and provide stochastic guarantees in the form of confidence intervals for the Reachtube bounds.  ...  We show that Neural ODEs, an emerging class of time-continuous neural networks, can be verified by solving a set of global-optimization problems.  ...  Acknowledgements The authors would like to thank the reviewers for their insightful comments. RH and RG were partially supported by Horizon-2020 ECSEL Project grant No. 783163 (iDev40).  ... 
doi:10.1609/aaai.v35i13.17372 fatcat:5i5gvkricvh3vjdc6xioddmqz4

GoTube: Scalable Statistical Verification of Continuous-Depth Models

Sophie A. Gruenbacher, Mathias Lechner, Ramin M. Hasani, Daniela Rus, Thomas A. Henzinger, Scott A. Smolka, Radu Grosu
2022 AAAI Conference on Artificial Intelligence  
We show that GoTube substantially outperforms state-of-theart verification tools in terms of the size of the initial ball, speed, time-horizon, task completion, and scalability on a large set of experiments  ...  We introduce a new statistical verification algorithm that formally quantifies the behavioral robustness of any timecontinuous process formulated as a continuous-depth model.  ...  GoTube is not yet suitable for the verification of stochastic dynamical systems, for instance, Neural Stochastic Differential Equations (Neural SDEs) Xu et al. 2021 ).  ... 
dblp:conf/aaai/GruenbacherLHRH22 fatcat:xurpydmw2feoviifppsez2d4yu

GoTube: Scalable Stochastic Verification of Continuous-Depth Models [article]

Sophie Gruenbacher, Mathias Lechner, Ramin Hasani, Daniela Rus, Thomas A. Henzinger, Scott Smolka, Radu Grosu
2021 arXiv   pre-print
We show that GoTube substantially outperforms state-of-the-art verification tools in terms of the size of the initial ball, speed, time-horizon, task completion, and scalability on a large set of experiments  ...  We introduce a new stochastic verification algorithm that formally quantifies the behavioral robustness of any time-continuous process formulated as a continuous-depth model.  ...  GoTube is not yet suitable for the verification of stochastic dynamical systems, for instance, Neural Stochastic Differential Equations (Neural SDEs) Xu et al. 2021) .  ... 
arXiv:2107.08467v2 fatcat:lird25izyjc6rb6qk3zaqw753y

Reachability Analysis of a General Class of Neural Ordinary Differential Equations [article]

Diego Manzanas Lopez, Patrick Musau, Nathaniel Hamilton, Taylor T. Johnson
2022 arXiv   pre-print
Specifically, our work extends an existing neural network verification tool to support neural ODEs.  ...  In this paper, we consider a general class of neural ODEs with varying architectures and layers, and introduce a novel reachability framework that allows for the formal analysis of their behavior.  ...  Any opinions, findings, and conclusions or recommendations expressed in this paper are those of the authors and do not necessarily reflect the views of AFOSR, DARPA, or NSF.  ... 
arXiv:2207.06531v1 fatcat:bsjgudckazaftcfne56hvgbjem

Scalable Synthesis of Verified Controllers in Deep Reinforcement Learning [article]

Zikang Xiong, Suresh Jagannathan
2021 arXiv   pre-print
Our key insight involves separating safety verification from neural controller, using pre-computed verified safety shields to constrain neural controller training which does not only focus on safety.  ...  Given the opacity and lack of interpretability of the neural policies that govern the behavior of such controllers, many existing approaches enforce safety properties through the use of shields, a dynamic  ...  Stochastic Linear Time-Variant Transition System We also study our verification algorithm on the stochastic linear time-variant transition system.  ... 
arXiv:2104.10219v2 fatcat:wmghro6mpzcmboj2ai5ihlplju

How to Learn a Model Checker [article]

Dung Phan, Radu Grosu, Nicola Paoletti, Scott A. Smolka, Scott D. Stoller
2017 arXiv   pre-print
Our method comprises a pipeline of analysis techniques for estimating and obtaining statistical guarantees on the classifier's prediction performance, as well as tuning techniques to improve such performance  ...  On these examples, we achieve an accuracy of 99.82% to 100% and a false-negative rate (incorrectly predicting that unsafe states are not reachable from a given state) of 0.0007 to 0.  ...  The work of [31] solves the verification problem for NNs with piecewiselinear activation function based on combining SAT solving with linear programming and on linear approximations of the network behavior  ... 
arXiv:1712.01935v1 fatcat:466w5f2y5fd6jam2q3birhlfpa

Stochastic Normalizing Flows [article]

Liam Hodgkinson, Chris van der Heide, Fred Roosta, Michael W. Mahoney
2020 arXiv   pre-print
Furthermore, by considering families of targeted SDEs with prescribed stationary distribution, we can apply VI to the optimization of hyperparameters in stochastic MCMC.  ...  Using the theory of rough paths, the underlying Brownian motion is treated as a latent variable and approximated, enabling efficient training of neural SDEs as random neural ordinary differential equations  ...  We would like to acknowledge DARPA, NSF, and ONR for providing partial support of this work.  ... 
arXiv:2002.09547v2 fatcat:5r3jktlfv5ffngpte4dygyk5tu

Verification for Machine Learning, Autonomy, and Neural Networks Survey [article]

Weiming Xiang and Patrick Musau and Ayana A. Wild and Diego Manzanas Lopez and Nathaniel Hamilton and Xiaodong Yang and Joel Rosenfeld and Taylor T. Johnson
2018 arXiv   pre-print
This survey presents an overview of verification techniques for autonomous systems, with a focus on safety-critical autonomous cyber-physical systems (CPS) and subcomponents thereof.  ...  Recently, the formal methods and formal verification community has developed methods to characterize behaviors in these LECs with eventual goals of formally verifying specifications for LECs, and this  ...  Future research will extend the usage to control MIMO systems with stochastic disturbance.  ... 
arXiv:1810.01989v1 fatcat:a5ax66lsxbho3fuxuh55ypnm6m

Learning Safe Neural Network Controllers with Barrier Certificates [article]

Hengjun Zhao, Xia Zeng, Taolue Chen, Zhiming Liu, Jim Woodcock
2020 arXiv   pre-print
We provide a novel approach to synthesize controllers for nonlinear continuous dynamical systems with control against safety properties. The controllers are based on neural networks (NNs).  ...  We train the controller-NN and barrier-NN simultaneously, achieving a verification-in-the-loop synthesis. We provide a prototype tool nncontroller with a number of case studies.  ...  Deshmukh for explanations on the bicycle model of Example 4.  ... 
arXiv:2009.09826v1 fatcat:chuykk5xn5hw7omlrwxo4ivrsi

Formal Certification Methods for Automated Vehicle Safety Assessment [article]

Tong Zhao, Ekim Yurtsever, Joel Paulson, Giorgio Rizzoni
2022 arXiv   pre-print
Recent advances in Level 3 and Level 4 autonomous driving have motivated more extensive study in safety guarantees of complicated AV maneuvers, which aligns with the goal of ISO 21448 (Safety of the Intended  ...  evolution of the system to provide guarantees.  ...  Li, “Efficient verification of control pp. 20–35, Springer Berlin Heidelberg, 2003. systems with neural network controllers,” in Proceedings of the 3rd [155] M. Althoff, D.  ... 
arXiv:2202.02818v2 fatcat:spgyrglbwjhshl2n43kyx722fa

NeuralExplorer: State Space Exploration of Closed Loop Control Systems Using Neural Networks [article]

Manish Goyal, Parasara Sridhar Duggirala
2020 arXiv   pre-print
We show how the approximation of sensitivity and inverse sensitivity can be used for computing estimates of the reachable set.  ...  We demonstrate the effectiveness of our approach by applying it not only to standard linear and nonlinear dynamical systems, but also to nonlinear hybrid systems and also neural network based feedback  ...  In: Proc. of the Workshop on Applied Verification for Continuous and Hybrid Systems (2015) 4.  ... 
arXiv:2007.05685v1 fatcat:3g4a37tbsjeyhiblhrcxqwunpa

Invertible Residual Networks [article]

Jens Behrmann, Will Grathwohl, Ricky T. Q. Chen, David Duvenaud, Jörn-Henrik Jacobsen
2019 arXiv   pre-print
Our empirical evaluation shows that invertible ResNets perform competitively with both state-of-the-art image classifiers and flow-based generative models, something that has not been previously achieved  ...  To compute likelihoods, we introduce a tractable approximation to the Jacobian log-determinant of a residual block.  ...  Acknowledgments We thank Rich Zemel for very helpful comments on an earlier version of the manuscript.  ... 
arXiv:1811.00995v3 fatcat:wnx5hz7rojcjrn2g2quroudobi

Deep Secure Encoding: An Application to Face Recognition [article]

Rohit Pandey, Yingbo Zhou, Venu Govindaraju
2015 arXiv   pre-print
The efficacy of the approach is shown on two face databases, namely, CMU-PIE and Extended Yale B, where we achieve state of the art matching performance, along with cancelability and high security with  ...  Furthermore, the scheme can work in both identification and verification modes.  ...  Since the code is binary, for the last layer of the neural network we apply the logistic function (f (x) = (1+exp −x ) −1 ) to get values between zero and one.  ... 
arXiv:1506.04340v1 fatcat:mmz3bkz2zvfgnnqetggjthyju4

Constrained Reinforcement Learning for Dynamic Optimization under Uncertainty

P. Petsagkourakis, I.O. Sandoval, E. Bradford, D. Zhang, E.A. del Rio-Chanona
2020 IFAC-PapersOnLine  
Backoffs are adjusted using the empirical cumulative distribution function to guarantee the satisfaction of a joint chance constraint.  ...  Many stochastic systems present the following obstacles: 1) plant-model mismatch, 2) process disturbances, 3) risks in violation of process constraints.  ...  However the focus here is on the probabilistic guarantee of the constraints.  ... 
doi:10.1016/j.ifacol.2020.12.361 fatcat:bznrotz3erbkpj2czfhjqpxuym
« Previous Showing results 1 — 15 out of 876 results