< Retour au sommaire
Tropical Algebra for Neural Network verification
Eric Goubault and Sylvie Putot le
Lieu: bâtiment 650, LISN
This talk is a joint talk with the Digicosme vrAI working
group.
In this talk, we will discuss range analyzes for feedforward neural networks,
which are basic primitives for applications such as robustness of neural
networks, compliance to specifications and reachability analysis of neural-
network feedback systems. Our approach focuses on ReLU (rectified linear unit)
feedforward neural nets that present specific difficulties: approaches that
exploit derivatives do not apply in general, the number of patterns of neuron
activations can be quite large even for small networks, and convex
approximations are generally too coarse. We use here set-based methods and
abstract interpretation that have been very successful in coping with similar
difficulties in classical program verification.
We are going to present an approach that abstracts ReLU feedforward neural
networks using tropical polyhedra. We show that tropical polyhedra can
efficiently abstract ReLU activation function, while being able to control the
loss of precision due to linear computations. We show how the connection between
ReLU networks and tropical rational functions can provide approaches for range
analysis of ReLU neural networks.