Learning Minimal Neural Specifications

TL;DR

This paper introduces methods to find minimal neural activation pattern specifications that improve neural network robustness verification, offering scalable insights into neuron contributions and significantly expanding verifiable bounds.

Contribution

The paper proposes three approaches to identify minimal neural activation pattern specifications, enhancing robustness verification and interpretability of neural networks.

Findings

Minimal NAP specifications use fewer neurons than previous methods.

Expanded verifiable bounds by several orders of magnitude.

Optimistic approach scales to large vision networks without verification tools.

Abstract

Formal verification is only as good as the specification of a system, which is also true for neural network verification. Existing specifications follow the paradigm of data as specification, where the local neighborhood around a reference data point is considered correct or robust. While these specifications provide a fair testbed for assessing model robustness, they are too restrictive for verifying any unseen test data points, a challenging task with significant real-world implications. Recent work shows great promise through a new paradigm, neural representation as specification, which uses neural activation patterns (NAPs) for this purpose. However, it computes the most refined NAPs, which include many redundant neurons. In this paper, we study the following problem: Given a neural network, find a minimal (general) NAP specification that is sufficient for formal verification of its robustness properties. Finding the minimal NAP specification not only expands verifiable bounds but also provides insights into which set of neurons contributes to the model's robustness. To address this problem, we propose three approaches: conservative, statistical, and optimistic. Each of these methods offers distinct strengths and trade-offs in terms of minimality and computational speed, making them suitable for scenarios with different priorities. Notably, the optimistic approach can probe potential causal links between neurons and the robustness of large vision neural networks without relying on verification tools, a task existing methods struggle to scale. Our experiments show that minimal NAP specifications use far fewer neurons than those from previous work while expanding verifiable boundaries by several orders of magnitude.