Penn Engineering

The School of Engineering and Applied Science, established in 1852, is composed of six academic departments and numerous interdisciplinary centers, institutes, and laboratories. At Penn Engineering, we are preparing the next generation of innovative engineers, entrepreneurs and leaders. Our unique culture of cooperation and teamwork, emphasis on research, and dedicated faculty advisors who teach as well as mentor, provide the ideal environment for the intellectual growth and development of well-rounded global citizens.

 

Filters

20201
1
Reset filters

Settings

Publication Type: Book ×

Search results

Now showing 1 - 1 of 1
  • Publication
    An Abstraction-Based Framework for Neural Network Verification
    (2020-01-01) Elboher, Yizhak Y; Gottschlich, Justin E; Katz, Guy
    Deep neural networks are increasingly being used as controllers for safety-critical systems. Because neural networks are opaque, certifying their correctness is a significant challenge. To address this issue, several neural network verification approaches have recently been proposed. However, these approaches afford limited scalability, and applying them to large networks can be challenging. In this paper, we propose a framework that can enhance neural network verification techniques by using over-approximation to reduce the size of the network—thus making it more amenable to verification. We perform the approximation such that if the property holds for the smaller (abstract) network, it holds for the original as well. The over-approximation may be too coarse, in which case the underlying verification tool might return a spurious counterexample. Under such conditions, we perform counterexample-guided refinement to adjust the approximation, and then repeat the process. Our approach is orthogonal to, and can be integrated with, many existing verification techniques. For evaluation purposes, we integrate it with the recently proposed Marabou framework, and observe a significant improvement in Marabou’s performance. Our experiments demonstrate the great potential of our approach for verifying larger neural networks.