An abstract domain for certifying neural networks
نویسندگان
چکیده
منابع مشابه
Certifying Domain-Specific Policies
Proof-checking code for compliance to safety policies potentially enables a product-oriented approach to certain aspects of software certification. To date, previous research has focused on generic, low-level programming-language properties such as memory type safety. In this paper we consider proof-checking higher-level domain-specific properties for compliance to safety policies. The paper fi...
متن کاملTWAM: A Certifying Abstract Machine for Logic Programs
Type-preserving (or typed) compilation uses typing derivations to certify correctness properties of compilation. We have designed and implemented a type-preserving compiler for a simply-typed dialect of Prolog we call T-Prolog. The crux of our approach is a new certifying abstract machine which we call the TypedWarren Abstract Machine (TWAM). The TWAMhas a dependent type system strong enough to...
متن کاملA Certifying Frontend for (Sub)polyhedral Abstract Domains
Convex polyhedra provide a relational abstraction of numerical properties for static analysis of programs by abstract interpretation. We describe a lightweight certification of polyhedral abstract domains using the Coq proof assistant. Our approach consists in delegating most computations to an untrusted backend and in checking its outputs with a certified frontend. The backend is free to imple...
متن کاملDomain-Adversarial Neural Networks
We introduce a new representation learning algorithm suited to the context of domain adaptation, in which data at training and test time come from similar but different distributions. Our algorithm is directly inspired by theory on domain adaptation suggesting that, for effective domain transfer to be achieved, predictions must be made based on a data representation that cannot discriminate bet...
متن کاملAn Abstract Domain for Multi-level Caches
Many program analyses and optimizations rely on knowledge of cache behavior. The precision of the underlying cache model is increasingly important with the recent uptake of multi-core and many-core architectures for two reasons. First, per-core cache sizes generally decrease as the number of cores becomes large resulting in more cache misses. Second, large scale sharing of the communication ban...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
ژورنال
عنوان ژورنال: Proceedings of the ACM on Programming Languages
سال: 2019
ISSN: 2475-1421
DOI: 10.1145/3290354