Search

Dynamic arrays, also referred to as vectors, are fundamental data structures used in many programs. Modeling their semantics efficiently is crucial when reasoning about such programs. The theory of arrays is widely supported but is not ideal, because the number of elements is fixed (determined by its index sort) and cannot be adjusted, which is a problem, given that the length of vectors often plays an important role when reasoning about vector programs. In this paper, we propose reasoning about vectors using a theory of sequences. We introduce the theory, propose a basic calculus adapted from one for the theory of strings, and extend it to efficiently handle common vector operations. We prove that our calculus is sound and show how to construct a model when it terminates with a saturated configuration. Finally, we describe an implementation of the calculus in cvc5 and demonstrate its efficacy by evaluating it on verification conditions for smart contracts and benchmarks derived from existing array benchmarks., Comment: IJCAR 2022

Noisy data, non-convex objectives, model misspecification, and numerical instability can all cause undesired behaviors in machine learning systems. As a result, detecting actual implementation errors can be extremely difficult. We demonstrate a methodology in which developers use an interactive proof assistant to both implement their system and to state a formal theorem defining what it means for their system to be correct. The process of proving this theorem interactively in the proof assistant exposes all implementation errors since any error in the program would cause the proof to fail. As a case study, we implement a new system, Certigrad, for optimizing over stochastic computation graphs, and we generate a formal (i.e. machine-checkable) proof that the gradients sampled by the system are unbiased estimates of the true mathematical gradients. We train a variational autoencoder using Certigrad and find the performance comparable to training the same model in TensorFlow., Comment: To appear at the Thirty-fourth International Conference on Machine Learning (ICML) 2017

The ability to deploy neural networks in real-world, safety-critical systems is severely limited by the presence of adversarial examples: slightly perturbed inputs that are misclassified by the network. In recent years, several techniques have been proposed for increasing robustness to adversarial examples --- and yet most of these have been quickly shown to be vulnerable to future attacks. For example, over half of the defenses proposed by papers accepted at ICLR 2018 have already been broken. We propose to address this difficulty through formal verification techniques. We show how to construct provably minimally distorted adversarial examples: given an arbitrary neural network and input sample, we can construct adversarial examples which we prove are of minimal distortion. Using this approach, we demonstrate that one of the recent ICLR defense proposals, adversarial retraining, provably succeeds at increasing the distortion required to construct adversarial examples by a factor of 4.2.

Standard Gibbs Free Energy of Formation MetaCyc. This file provides standard Gibbs Free Energy of Formation taken from MetaCyc database [29] for metabolites used in the simulation example in Results section. (PDF 51.4 kb)

Derivation for steady state rate equation. This file provides a detailed derivation for steady state rate equation of a single reactant and single product reversible metabolic reaction presented in Methods section. (PDF 92.8 kb)

Equilibrium constant K eq . The equilibrium constant K eq is assumed to be a known constant. This file provides the calculation of equilibrium constant K eq based on standard Gibbs Free Energies of Formation. (PDF 98.8 kb)

Symbolic Model Checking extends the scope of verification algorithms that can be handled automatically, by using symbolic representations rather than explicitly searching the entire state space of the model. However even the most sophisticated symbolic methods cannot be directly applied to many of today's large designs because of the state explosion problem. Approximate symbolic model checking is an attempt to trade off accuracy with the capacity to deal with bigger designs. This paper explores the idea of using overlapping projections as the underlying approximation scheme. The idea is evaluated by applying it to several modules from the I/O unit in the Stanford FLASH Multiprocessor, and some larger circuits in ISCAS89 benchmark suite.

Computer Science Department

The massification of higher education, the rapid development of new academic subjects and fields, the growing international market competition among universities, the associated deregulation of government policy and growing institutional autonomy, the commercial provision of quality information, and the resulting academic arms race for research reputation and prestige have challenged the traditional ways of maintaining academic standards. In this concluding chapter we reflect on the lessons to be learned from our studies of professional regulation, market regulation, and state regulation of academic quality and we explore how the necessary balance among the different forces can best be accomplished. We also attempt to derive from these studies of individual instruments some general guidelines that may prove useful in designing national framework conditions for assuring academic standards in the university sector.

University rankings have become widely influential in the last 10 years, both on a national and international scale. Rankings as a consumer information tool can function as an effective quality assurance mechanism. Most existing university rankings, however, seem to distort rather than improve the higher education market. The CHE ranking in Germany is an exception. It is a carefully designed ranking that minimizes the main conceptual and methodological problems that university rankings commonly face. The analysis in the chapter concludes that commercially oriented entities alone cannot provide a high quality university ranking. Original data collection and data verification is a costly activity and there is a strong incentive for commercial providers to rely on easily available statistics. Therefore, even if a commercial venue can be effective in compiling, presenting and marketing relevant information, the quality of a university ranking depends on the data collected by public or not-for-profit agencies.

We present a method for automatically verifying asynchronous sequential circuits using temporal logic specifications.

The temporal logic model algorithm of E.M. Clarke et al. (ACM Trans. Prog. Lang. Syst., vol.8, no.2, p.244-63, 1986) is modified to represent a state graph using binary decision diagrams (BDDs). Because this representation captures some of the regularity in the state space of sequential circuits with data path logic, one is able to verify circuits with an extremely large number of states. This new technique is demonstrated on a synchronous pipelined design with approximately 5×1020 states. The logic that is used to specify circuits is a propositional temporal logic of branching time, called CTL or Computation Tree Logic. The model checking algorithm handles full CTL with fairness constraints. Consequently. it is possible to handle a number of important liveness and fairness properties. which would otherwise not be expressible in CTL. The method presented is not necessarily a replacement for brute-force state-enumeration methods but an alternative that may work efficiently when the brute force methods fail