Variational Autoencoders (VAEs) have become increasingly popular and deployed in safety-critical applications. In such applications, we want to give certified probabilistic guarantees on performance under adversarial attacks. We propose a novel method, CIVET, for certified training of VAEs. CIVET depends on the key insight that we can bound worst-case VAE error by bounding the error on carefully chosen support sets at the latent layer. We show this point mathematically and present a novel training algorithm utilizing this insight. We show in an extensive evaluation across different datasets (in both the wireless and vision application areas), architectures, and perturbation magnitudes that our method outperforms SOTA methods achieving good standard performance with strong robustness guarantees.

Universal Adversarial Perturbations (UAPs) are imperceptible, image-agnostic vectors that cause deep neural networks (DNNs) to misclassify inputs with high probability. In practical attack scenarios, adversarial perturbations may undergo transformations such as changes in pixel intensity, scaling, etc. before being added to DNN inputs. Existing methods do not create UAPs robust to these real-world transformations, thereby limiting their applicability in practical attack scenarios. In this work, we introduce and formulate UAPs robust against real-world transformations. We build an iterative algorithm using probabilistic robustness bounds and construct UAPs robust to transformations generated by composing arbitrary sub-differentiable transformation functions. We perform an extensive evaluation on the popular CIFAR-10 and ILSVRC 2012 datasets measuring our UAPs’ robustness under a wide range common, real-world transformations such as rotation, contrast changes, etc. We further show that by using a set of primitive transformations our method generalizes well to unseen transformations such as fog, JPEG compression, etc. Our results show that our method can generate UAPs up to 23% more robust than state-of-the-art baselines.

Existing mobile edge computing (MEC) systems are facing the challenges of limited resources and highly dynamic network environments. How to allocate resources to maintain the efficiency and timeliness of data and tasks is still an open issue. To address this problem, we propose a novel framework for UAV-assisted MEC systems using federated multi-agent reinforcement learning. First, we formulate a joint optimization problem as a multi-agent Markov decision process by jointly minimizing the average age of information and maximizing the number of recent tasks. Second, we design a novel scheduling algorithm for online collaborative resources by adopting multiple agents to learn and make decisions in accordance with the overall interests through federal learning. Finally, an experience replay mechanism for the internal experience pool is introduced to further improve learning efficiency. Experimental results show that our proposed algorithm is superior to the recent typical reinforcement learning-based algorithms. It not only has higher efficiency in task processing and data freshness, but also has more stable performance and adaptability across diverse experimental conditions.

We consider the verification of input-relational properties defined over deep neural networks (DNNs) such as robustness against universal adversarial perturbations, monotonicity, etc. Precise verification of these properties requires reasoning about multiple executions of the same DNN. We introduce a novel concept of difference tracking to compute the difference between the outputs of two executions of the same DNN at all layers. We design a new abstract domain, DiffPoly for efficient difference tracking that can scale large DNNs. DiffPoly is equipped with custom abstract transformers for common activation functions (ReLU, Tanh, Sigmoid, etc.) and affine layers and can create precise linear cross-execution constraints. We implement an input-relational verifier for DNNs called RaVeN which uses DiffPoly and linear program formulations to handle a wide range of input-relational properties. Our experimental results on challenging benchmarks show that by leveraging precise linear constraints defined over multiple executions of the DNN, RaVeN gains substantial precision over baselines on a wide range of datasets, networks, and input-relational properties.

In this work, we propose a deep reinforcement learning (DRL) algorithm DoubleJump which can master the game of Amazons efficiently. To address the bottleneck problem of sparse supervision signal in DRL, we split the neural network into rule network and skill network, using huge amounts of inexpensive data with game rule information and scarce data containing game skill information to train two networks respectively. Besides, we split the three sub-actions of each action into independent states during Monte-Carlo tree search (MCTS), to improve the probability of finding the global optimal state and reduce the average branching factor. The experimental results show our algorithm reaches about 130:70 in the zero-knowledge learning compared with the AlphaGo Zero algorithm, significantly improves the learning speed, and then alleviates the severe dependence on computing resources.

A program is said to have a determinacy race if logically parallel parts of a program access the same memory location and one of the accesses is a write. These races are generally bugs in the program since they lead to non-deterministic program behavior — different schedules of the program can lead to different results. Most prior work on detecting these races focuses on a subclass of programs with series-parallel or nested parallelism. This paper presents a race-detection algorithm for detecting races in a more general class of programs, namely programs that include arbitrary ordering constraints in additional to the series-parallel constructs. The algorithm performs a serial execution of the program, augmented to detect races, in O(T1 + k2) time, where T1 is the sequential running time of the original program and k is the number of non series-parallel constraints. The main technical novelty of this paper is a new data structure, R-Sketch, for answering reachability queries in nearly series-parallel (SP) directed acyclic graphs (DAGs). Given as input a graph comprising an n-node series parallel graph and k additional non-SP edges, the total construction time of the data structure is O(n + k2), and each reachability query can be answered in O(1) time. The data structure is traversally incremental, meaning that it supports the insertion of nodes/edges, but only as they are discovered through a graph traversal.

New media platforms, including the network community, microblogging, WeChat, live show platform, have become the main platforms and channels for information dissemination and diffusion, at the same time, these platforms are also favored by the majority of merchants, and how to use these new media platforms for precise marketing is an urgent problem. This paper proposes a precise marketing algorithm based on topic community discovery, which can use the link relation between user and user and the user generated content to determine the user's community division, then the merchants can put different advertising content for different topic community. The algorithm uses the LDA algorithm to analyze the text information in the user-generated content, and analyzes the link relation between the user and the user through the classical community discovery method, and combines the community structure to discover the network structure. These topic communities are the basis of network precise marketing. Extensive experiments are done using real world data, which prove that the algorithm proposed in this paper can effectively use the user topic and user link relationship to find the topic community, thus providing the basis for precise marketing.

In order to increase throughput and improve system performance for free space optical communications, we consider a pilot-symbol assisted modulation (PSAM) subcarrier phase shift keying (PSK) system with maximum likelihood (ML) phase estimation. Using the phase error probability density function of ML estimation phase error, we study the error rate performance of the subcarrier PSK system with carrier phase error in lognormal turbulence channels. Results reveal the differences between the imperfectly synchronized system and the perfectly synchronized system, and how the performances are influenced by the pilot symbol length. An analytical expression of asymptotic signal-to-noise ratio penalty is derived and verified with numerical results.