16 January 2023 in ACM Computing Surveys
Recent years have witnessed the widespread popularity of Internet of things (IoT). By providing sufficient data for model training and inference, IoT has promoted the development of artificial intelligence (AI) to a great extent. Under this background and trend, the traditional cloud computing model may nevertheless encounter many problems in independently tackling the massive data generated by IoT and meeting corresponding practical needs. In response, a new computing model called edge computing (EC) has drawn extensive attention from both industry and academia. With the continuous deepening of the research on EC, however, scholars have found that traditional (non-AI) methods have their limitations in enhancing the performance of EC. Seeing the successful application of AI in various fields, EC researchers start to set their sights on AI, especially from a perspective of machine learning, a branch of AI that has gained increased popularity in the past decades. In this article, we first explain the formal definition of EC and the reasons why EC has become a favorable computing model. Then, we discuss the problems of interest in EC. We summarize the traditional solutions and hightlight their limitations. By explaining the research results of using AI to optimize EC and applying AI to other fields under the EC architecture, this article can serve as a guide to explore new research ideas in these two aspects while enjoying the mutually beneficial relationship between AI and EC.
8 September 2021 in IEEE Transactions on Sustainable Energy
With the increasing concern on climate change and global warming, the reduction of carbon emission becomes an important topic in many aspects of human society. The development of energy Internet (EI) makes it possible to achieve better utilization of distributed renewable energy sources with the power sharing functionality introduced by energy routers (ERs). In this paper, a bottom-up EI architecture is designed, and a novel data-driven dynamical control strategy is proposed. Intelligent controllers augmented by deep reinforcement learning (DRL) techniques are adopted for the operation of each microgrid independently in the bottom layer. Moreover, the concept of curriculum learning (CL) is integrated into DRL to improve the sample efficiency and accelerate the training process. Based on the power exchange plan determined in the bottom layer, considering the stochastic nature of electricity price in the future power market, the optimal power dispatching scheme in the upper layer is decided via model predictive control. The simulation has shown that, under the bottom-up architecture, compared with the conventional methods such as proportional integral and optimal power flow, the proposed method reduces overall generation cost by 7.1% and 37%, respectively. Meanwhile, the introduced CL-based training strategy can significantly speed up the convergence during the training of DRL. Last but not least, our method increases the profit of energy trading between ERs and the main grid.
14 December 2018 in IEEE Transactions on Image Processing
Salient segmentation aims to segment out attentiongrabbing regions, a critical yet challenging task and the foundation of many high-level computer vision applications. It requires semantic-aware grouping of pixels into salient regions and benefits from the utilization of global multi-scale contexts to achieve good local reasoning. Previous works often address it as two-class segmentation problems utilizing complicated multi-step procedures including refinement networks and complex graphical models. We argue that semantic salient segmentation can instead be effectively resolved by reformulating it as a simple yet intuitive pixel-pair based connectivity prediction task. Following the intuition that salient objects can be naturally grouped via semanticaware connectivity between neighboring pixels, we propose a pure Connectivity Net (ConnNet). ConnNet predicts connectivity probabilities of each pixel with its neighboring pixels by leveraging multi-level cascade contexts embedded in the image and long-range pixel relations. We investigate our approach on two tasks, namely salient object segmentation and salient instancelevel segmentation, and illustrate that consistent improvements can be obtained by modeling these tasks as connectivity instead of binary segmentation tasks for a variety of network architectures. We achieve state-of-the-art performance, outperforming or being comparable to existing approaches while reducing inference time due to our less complex approach.
20 May 2024 in IEEE Transactions on Big Data
Federated learning (FL) is a machine learning paradigm, which enables multiple and decentralized clients to collaboratively train a model under the orchestration of a central aggregator. FL can be a scalable machine learning solution in Big Data scenarios. Traditional FL relies on the trust assumption of the central aggregator, which forms cohorts of clients honestly. However, a malicious aggregator, in reality, could abandon and replace the client's training models, or insert fake clients, to manipulate the final training results. In this work, we introduce $ \tt {zkFL}$, which leverages zero-knowledge proofs to tackle the issue of a malicious aggregator during the training model aggregation process. To guarantee the correct aggregation results, the aggregator provides a proof per round, demonstrating to the clients that the aggregator executes the intended behavior faithfully. To further reduce the verification cost of clients, we use blockchain to handle the proof in a zero-knowledge way, where miners (i.e., the participants validating and maintaining the blockchain data) can verify the proof without knowing the clients’ local and aggregated models. The theoretical analysis and empirical results show that $ \tt {zkFL}$ achieves better security and privacy than traditional FL, without modifying the underlying FL network structure or heavily compromising the training speed.