Simplifying Neural Networks using Formal Verification
This work addresses the need for more efficient and reliable neural networks in safety-critical domains like aviation, though it is incremental as it builds on existing verification engines.
The authors tackled the problem of simplifying neural networks by using formal verification to reduce network size without harming accuracy, achieving up to a 10% reduction in size on real-world DNNs for aircraft collision avoidance.
Deep neural network (DNN) verification is an emerging field, with diverse verification engines quickly becoming available. Demonstrating the effectiveness of these engines on real-world DNNs is an important step towards their wider adoption. We present a tool that can leverage existing verification engines in performing a novel application: neural network simplification, through the reduction of the size of a DNN without harming its accuracy. We report on the work-flow of the simplification process, and demonstrate its potential significance and applicability on a family of real-world DNNs for aircraft collision avoidance, whose sizes we were able to reduce by as much as 10%.