Reachability In Simple Neural Networks
This work addresses the fundamental verification problem for neural networks, showing inherent computational hardness even in simple cases, which is important for researchers and practitioners in AI safety and formal methods.
The paper investigates the computational complexity of the reachability problem in neural networks, showing that NP-hardness holds even for highly restricted networks with a single hidden layer, output dimension of one, and simple weight/bias constraints.
We investigate the complexity of the reachability problem for (deep) neural networks: does it compute valid output given some valid input? It was recently claimed that the problem is NP-complete for general neural networks and specifications over the input/output dimension given by conjunctions of linear inequalities. We recapitulate the proof and repair some flaws in the original upper and lower bound proofs. Motivated by the general result, we show that NP-hardness already holds for restricted classes of simple specifications and neural networks. Allowing for a single hidden layer and an output dimension of one as well as neural networks with just one negative, zero and one positive weight or bias is sufficient to ensure NP-hardness. Additionally, we give a thorough discussion and outlook of possible extensions for this direction of research on neural network verification.