Speeding up neural network robustness verification via algorithm configuration and an optimised mixed integer linear programming solver portfolio
نویسندگان
چکیده
Abstract Despite their great success in recent years, neural networks have been found to be vulnerable adversarial attacks. These attacks are often based on slight perturbations of given inputs that cause them misclassified. Several methods proposed formally prove robustness a network against such However, these typically give rise high computational demands, which severely limit scalability. Recent state-of-the-art approaches state the verification task as minimisation problem, is formulated and solved mixed-integer linear programming (MIP) problem. We extend this approach by leveraging automated algorithm configuration techniques and, more specifically, construct portfolio MIP solver configurations optimised for task. test two recent, MIP-based engines, $$\mathrm {MIPVerify}$$ MIPVerify {Venus}$$ Venus , achieve substantial improvements CPU time average factors up 4.7 10.3, respectively.
منابع مشابه
An evolutionary solver for mixed integer programming
In this paper we introduce an evolutionary algorithm for the solution of mixed integer programs. The strategy is based on the separation of the set of variables into the integer subset and the continuous subset. The main idea is that if the integer variables are fixed by the evolutionary system, the continuous ones can be determined in function of them by a linear program, which simultaneously ...
متن کاملAn Exact Rational Mixed-Integer Programming Solver
We present an exact rational solver for mixed-integer linear programming which avoids the numerical inaccuracies inherent in the floating-point computations adopted in existing software. This allows the solver to be used for establishing fundamental theoretical results and in applications where correct solutions are critical due to legal and financial consequences. Our solver is a hybrid symbol...
متن کاملAn evolutionary solver for linear integer programming
In this paper we introduce an evolutionary algorithm for the solution of linear integer programs. The strategy is based on the separation of the variables into the integer subset and the continuous subset; the integer variables are fixed by the evolutionary system, and the continuous ones are determined in function of them, by a linear program solver. We report results obtained for some standar...
متن کاملA mixed integer nonlinear programming model for optimizing a gas pipeline transmission linear network
The technical equipment developed and used in both installation and operation processes in refineries, oil and gas pipelines, and gas booster stations has always been expensive. Hence, managers at different organizational levels are keen to find methods to control and reduce these costs. Generally speaking, the operators in a gas booster station choose the operating devices without considering ...
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: Machine Learning
سال: 2022
ISSN: ['0885-6125', '1573-0565']
DOI: https://doi.org/10.1007/s10994-022-06212-w