Neural networks (NNs) playing the role of controllers have demonstrated impressive empirical performance on challenging control problems. However, the potential adoption of NN controllers in real-life applications has been significantly impeded by the growing concerns over the safety of these NN-controlled systems (NNCSs). In this work, we present POLAR-Express, an efficient and precise formal reachability analysis tool for verifying the safety of NNCSs. POLAR-Express uses Taylor model (TM) arithmetic to propagate TMs layer-by-layer across an NN to compute an overapproximation of the NN. It can be applied to analyze any feedforward NNs with continuous activation functions, such as ReLU, Sigmoid, and Tanh activation functions that cover the common benchmarks for NNCS reachability analysis. Compared with its earlier prototype POLAR, we develop a novel approach in POLAR-Express to propagate TMs more efficiently and precisely across ReLU activation functions, and provide parallel computation support for TM propagation, thus significantly improving the efficiency and scalability. Across the comparison with six other state-of-the-art tools on a diverse set of common benchmarks, POLAR-Express achieves the best verification efficiency and tightness in the reachable set analysis. POLAR-Express is publicly available at https://github.com/ChaoHuang2018/POLAR_Tool .
Model-based reinforcement learning has been widely studied for controller synthesis in cyber-physical systems (CPSs). In particular, for safety-critical CPSs, it is important to formally certify system properties (e.g., safety, stability) under the learned RL controller. However, as existing methods typically conduct formal verification after the controller has been learned, it is often difficult to obtain any certificate, even after many iterations between learning and verification. To address this challenge, we propose a framework that jointly conducts reinforcement learning and formal verification by formulating and solving a novel bilevel optimization problem, which is end-to-end differentiable by the gradients from the value function and certificates formulated by linear programs and semi-definite programs. In experiments, our framework is compared with a baseline model-based stochastic value gradient (SVG) method and its extension to solve constrained Markov Decision Processes (CMDPs) for safety. The results demonstrate the significant advantages of our framework in finding feasible controllers with certificates, i.e., Barrier functions and Lyapunov functions that formally ensure system safety and stability, available on Github.
We present REGLO, a novel methodology for repairing pretrained neural networks to satisfy global robustness and individual fairness properties. A neural network is said to be globally robust with respect to a given input region if and only if all the input points in the region are locally robust. This notion of global robustness also captures the notion of individual fairness as a special case. We prove that any counterexample to a global robustness property must exhibit a corresponding large gradient. For ReLU networks, this result allows us to efficiently identify the linear regions that violate a given global robustness property. By formulating and solving a suitable robust convex optimization problem, REGLO then computes a minimal weight change that will provably repair these violating linear regions.
The development of the recurrent neural network (RNN) in machine translation (MT) has greatly improved its translation speed and accuracy. Although a larger proportion of the discourse power of MT is increasingly taken in translation activities, MT cannot monopolize human translation due to the probability algorithm. There are two reasons. First, the order and format of some texts will affect the accuracy of machine translation according to the algorithm of probability theory. Second, “memory” is not only an indispensable core technique in machine translation but also an insurmountable component in translation psychology. Although the current research on MT based on RNNs has made some achievements, it is still impossible to obtain the most scientific combination mode of the parts of speech according to probabilistic computing; therefore, the machine itself cannot distinguish the correctness of the source language, which leads to the fact that the input of the source language with the wrong structure may still obtain the correct translation. This paper explores possible ways to optimize MT by researching the internal relations among probability theory, translation psychology and MT.
Data quantity and quality are crucial factors for data-driven learning methods. In some target problem domains, there are not many data samples available, which could significantly hinder the learning process. While data from similar domains may be leveraged to help through domain adaptation, obtaining high-quality labeled data for those source domains themselves could be difficult or costly. To address such challenges on data insufficiency for classification problem in a target domain, we propose a weak adaptation learning (WAL) approach that leverages unlabeled data from a similar source domain, a low-cost weak annotator that produces labels based on task-specific heuristics, labeling rules, or other methods (albeit with inaccuracy), and a small amount of labeled data in the target domain. Our approach first conducts a theoretical analysis on the error bound of the trained classifier with respect to the data quantity and the performance of the weak annotator, and then introduces a multi-stage weak adaptation learning method to learn an accurate classifier by lowering the error bound. Our experiments demonstrate the effectiveness of our approach in learning an accurate classifier with limited labeled data in the target domain and unlabeled data in the source domain.