AI RESEARCH
The Complexity of Verifying Feedforward Neural Networks in Quantised Settings
arXiv CS.LG
•
ArXi:2605.29537v1 Announce Type: cross We investigate the computational complexity of neural network verification in quantised settings. We distinguish three classes of Feedforward Neural Networks (FNNs): rational FNNs with exact rational weights, quantised FNNs whose weights come from a finite-width arithmetic, and dynamically quantised FNNs in which rational networks are evaluated with respect to a given finite-width arithmetic. We consider two types of specifications used in the literature.