A domain-theoretic framework for robustness analysis of neural networks
From MaRDI portal
Publication:6149907
DOI10.1017/S0960129523000142arXiv2203.00295OpenAlexW4377822310MaRDI QIDQ6149907FDOQ6149907
Authors: Can Zhou, Yiran Li, Amin Farjudian 
Publication date: 5 March 2024
Published in: Mathematical Structures in Computer Science (Search for Journal in Brave)
Abstract: A domain-theoretic framework is presented for validated robustness analysis of neural networks. First, global robustness of a general class of networks is analyzed. Then, using the fact that Edalat's domain-theoretic L-derivative coincides with Clarke's generalized gradient, the framework is extended for attack-agnostic local robustness analysis. The proposed framework is ideal for designing algorithms which are correct by construction. This claim is exemplified by developing a validated algorithm for estimation of Lipschitz constant of feedforward regressors. The completeness of the algorithm is proved over differentiable networks, and also over general position ReLU networks. Computability results are obtained within the framework of effectively given domains. Using the proposed domain model, differentiable and non-differentiable networks can be analyzed uniformly. The validated algorithm is implemented using arbitrary-precision interval arithmetic, and the results of some experiments are presented. The software implementation is truly validated, as it handles floating-point errors as well.
Full work available at URL: https://arxiv.org/abs/2203.00295
Cites Work
- SHRAD: A language for sequential real number computation
- Extreme value theory. An introduction.
- Title not available (Why is that?)
- Title not available (Why is that?)
- A Domain-Theoretic Account of Picard's Theorem
- Introduction to Interval Analysis
- Title not available (Why is that?)
- Title not available (Why is that?)
- Title not available (Why is that?)
- Title not available (Why is that?)
- Continuous Lattices and Domains
- Properly injective spaces and function spaces
- Title not available (Why is that?)
- The way-below relation of function spaces over semantic domains
- Non-Hausdorff topology and domain theory. Selected topics in point-set topology
- Motivations for an arbitrary precision interval arithmetic and the MPFI library
- Denotational semantics of hybrid automata
- Safe \& robust reachability analysis of hybrid systems
- Concrete models of computation for topological algebras
- Dynamical systems, measures, and fractals via domain theory
- Domain theory and differential calculus (functions of one variable)
- A continuous derivative for real-valued functions
- Effectively given domains
- A domain-theoretic approach to computability on the real line
- Domains for Computation in Mathematics, Physics and Exact Real Arithmetic
- Strictness analysis for higher-order functions
- Title not available (Why is that?)
- Abstract Interpretation, Logical Relations, and Kan Extensions
- Real number computability and domain theory
- Integration in Real PCF
- Title not available (Why is that?)
- Benign overfitting in linear regression
- A language for differentiable functions
- Domain theoretic second-order Euler's method for solving initial value problems
- System analysis and robustness
- Exploiting verified neural networks via floating point numerical error
- Abstract neural networks
- Clarke's generalized gradient and Edalat's L-derivative
- Semantics of query-driven communication of exact values
- Compositional semantics of dataflow networks with query-driven communication of exact values
This page was built for publication: A domain-theoretic framework for robustness analysis of neural networks
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6149907)
