• Corpus ID: 239009452

Sound and Complete Neural Network Repair with Minimality and Locality Guarantees

@article{Fu2021SoundAC,
  title={Sound and Complete Neural Network Repair with Minimality and Locality Guarantees},
  author={Feisi Fu and Wenchao Li},
  journal={ArXiv},
  year={2021},
  volume={abs/2110.07682}
}
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… 

Figures and Tables from this paper

References

SHOWING 1-10 OF 32 REFERENCES
Provable repair of deep neural networks
TLDR
The Provable Repair problem is introduced, which is the problem of repairing a network N to construct a new network N′ that satisfies a given specification, and the introduction of a Decoupled DNN architecture, which allows for provable repair to a linear programming problem.
Minimal Modifications of Deep Neural Networks using Verification
TLDR
This work uses recent advances in DNN verification and proposes a technique for modifying a DNN according to certain requirements, in a way that is provably minimal, does not require any retraining, and is thus less likely to affect other aspects of the DNN’s behavior.
ImageNet classification with deep convolutional neural networks
TLDR
A large, deep convolutional neural network was trained to classify the 1.2 million high-resolution images in the ImageNet LSVRC-2010 contest into the 1000 different classes and employed a recently developed regularization method called "dropout" that proved to be very effective.
Basic Properties of Convex Polytopes
TLDR
This chapter tries to give a short introduction, provide a sketch of “what polytopes look like” and “how they behave,” with many explicit examples, and briefly state some main results (where further details are in the subsequent chapters of this Handbook).
End to end learning for self-driving
  • cars. CoRR,
  • 2016
Towards Repairing Neural Networks Correctly
TLDR
This work proposes a runtime repairing method to ensure the correctness of neural networks within certain input regions and generates neural networks which are guaranteed to satisfy the properties, whilst being consistent with the original neural network most of the time.
DISCO Verification: Division of Input Space into COnvex polytopes for neural network verification
TLDR
This paper proposes a method to simplify the verification problem by operating a partitionning into multiple linear subproblems, and performs an empirical analysis of neural networks to estimate the number of linear regions, and compares them to the bounds currently known.
Few-Shot Guided Mix for DNN Repairing
TLDR
A novel augmentation-based repairing method for deep neural networks that automatically learns to resolve the potential failure patterns in the DNN operational environment, and is able to achieve high generalization to the collected failure examples and other similar suspicious data.
Editable Neural Networks
TLDR
This work proposes Editable Training, a model-agnostic training technique that encourages fast editing of the trained model, and empirically demonstrates the effectiveness of this method on large-scale image classification and machine translation tasks.
Guaranteeing Safety for Neural Network-Based Aircraft Collision Avoidance Systems
TLDR
A method to provide safety guarantees when using a neural network collision avoidance system is proposed and experiments with systems inspired by ACAS X show that neural networks giving either horizontal or vertical maneuvers can be proven safe.
...
...