A Tensor-Based Compiler and a Runtime for Neuron-Level DNN Certifier Specifications

2025-07-26

AI Summary

Existing DNN abstract interpretation-based verifiers suffer from a semantic gap between design (neuron-level abstractions) and implementation (tensor-level operations), resulting in high development overhead and poor scalability. To bridge this gap, we propose the first compiler framework tailored for verification: it introduces a stack-based intermediate representation and shape-aware analysis to automatically lift neuron-level semantics to tensor operations; designs the g-BCSR dual sparse format, specifically optimized for inherent sparsity patterns in DNNs; and integrates domain-specific optimization rewrite rules with runtime shape inference to enable efficient, formally correct layer-wise computation. The framework significantly lowers the barrier to developing new verifiers—achieving performance on par with hand-optimized implementations—while supporting flexible, scalable verification across diverse DNN architectures.

Technology Category

• Machine Learning: Hardware-aware ML
• Computer Vision: Learning & Optimization for CV
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

The uninterpretability of DNNs has led to the adoption of abstract interpretation-based certification as a practical means to establish trust in real-world systems that rely on DNNs. However, the current landscape supports only a limited set of certifiers, and developing new ones or modifying existing ones for different applications remains difficult. This is because the mathematical design of certifiers is expressed at the neuron level, while their implementations are optimized and executed at the tensor level. This mismatch creates a semantic gap between design and implementation, making manual bridging both complex and expertise-intensive -- requiring deep knowledge in formal methods, high-performance computing, etc. We propose a compiler framework that automatically translates neuron-level specifications of DNN certifiers into tensor-based, layer-level implementations. This is enabled by two key innovations: a novel stack-based intermediate representation (IR) and a shape analysis that infers the implicit tensor operations needed to simulate the neuron-level semantics. During lifting, the shape analysis creates tensors in the minimal shape required to perform the corresponding operations. The IR also enables domain-specific optimizations as rewrites. At runtime, the resulting tensor computations exhibit sparsity tied to the DNN architecture. This sparsity does not align well with existing formats. To address this, we introduce g-BCSR, a double-compression format that represents tensors as collections of blocks of varying sizes, each possibly internally sparse. Using our compiler and g-BCSR, we make it easy to develop new certifiers and analyze their utility across diverse DNNs. Despite its flexibility, the compiler achieves performance comparable to hand-optimized implementations.

Problem

• Bridging neuron-level DNN certifier design to tensor-level implementation
• Automating translation of certifier specs to optimized tensor code
• Enabling efficient sparse tensor computation for DNN certification

Innovation

• Compiler translates neuron-level to tensor-level
• Stack-based IR and shape analysis enable lifting
• g-BCSR format handles sparse tensor computations