AI Navigate
インサイト最新記事一覧AI大全

The Luna Bound Propagator for Formal Analysis of Neural Networks

arXiv cs.LG / 2026/3/26

📰 ニュースIdeas & Deep AnalysisTools & Practical UsageModels & Research
原文を読む →
共有:

要点

  • The paper introduces Luna, a new C++ bound propagator aimed at formal neural network verification, addressing alpha-CROWN’s prior limitation to Python implementations.
  • Luna supports multiple verification-relevant techniques—Interval Bound Propagation, CROWN, and alpha-CROWN—while operating over a general computational graph.
  • The authors present Luna’s architecture and evaluate it using VNN-COMP 2025 benchmarks, focusing on bound tightness and computational efficiency.
  • Results indicate Luna is competitive with the state-of-the-art alpha-CROWN implementation in both effectiveness and runtime performance, suggesting practical viability for production DNN verifiers.

Abstract

The parameterized CROWN analysis, a.k.a., alpha-CROWN, has emerged as a practically successful bound propagation method for neural network verification. However, existing implementations of alpha-CROWN are limited to Python, which complicates integration into existing DNN verifiers and long-term production-level systems. We introduce Luna, a new bound propagator implemented in C++. Luna supports Interval Bound Propagation, the CROWN analysis, and the alpha-CROWN analysis over a general computational graph. We describe the architecture of Luna and show that it is competitive with the state-of-the-art alpha-CROWN implementation in terms of both bound tightness and computational efficiency on benchmarks from VNN-COMP 2025.

関連おすすめサービス

※当サイトはアフィリエイト広告を利用しています
🎧

Notta搭載AI議事録イヤホン ZENCHORD1

AI時代の仕事術。Notta搭載で会議の議事録を自動生成するスマートイヤホン。

🎙️

AI搭載ボイスレコーダー Plaud

世界100万人が愛用。AIで文字起こし・要約を自動化するボイスレコーダー。

🖼️

画像高画質化AIツール Aiarty Image Enhancer

AIで画像を高画質化。写真・イラストを簡単にアップスケール。