Sound and Complete Neural Network Repair with Minimality and Locality Guarantees
This work addresses the challenge of safely correcting neural network errors for applications requiring reliability, though it is incremental as it builds on existing repair techniques.
The paper tackles the problem of repairing neural networks with ReLU activations by introducing a method that applies localized changes to fix buggy inputs while minimizing global alterations, resulting in significant improvements in locality and reduced negative side effects compared to existing methods.
We present a novel methodology for repairing neural networks that use ReLU activation functions. Unlike existing methods that rely on modifying the weights of a neural network which can induce a global change in the function space, our approach applies only a localized change in the function space while still guaranteeing the removal of the buggy behavior. By leveraging the piecewise linear nature of ReLU networks, our approach can efficiently construct a patch network tailored to the linear region where the buggy input resides, which when combined with the original network, provably corrects the behavior on the buggy input. Our method is both sound and complete -- the repaired network is guaranteed to fix the buggy input, and a patch is guaranteed to be found for any buggy input. Moreover, our approach preserves the continuous piecewise linear nature of ReLU networks, automatically generalizes the repair to all the points including other undetected buggy inputs inside the repair region, is minimal in terms of changes in the function space, and guarantees that outputs on inputs away from the repair region are unaltered. On several benchmarks, we show that our approach significantly outperforms existing methods in terms of locality and limiting negative side effects. Our code is available on GitHub: https://github.com/BU-DEPEND-Lab/REASSURE.