Large language models (LLMs) have emerged as a new paradigm for Text-to-SQL task. However, the absence of a systematical benchmark inhibits the development of designing effective, efficient and economic LLM-based Text-to-SQL solutions. To address this challenge, in this paper, we first conduct a systematical and extensive comparison over existing prompt engineering methods, including question representation, example selection and example organization, and with these experimental results, we elaborate their pros and cons. Based on these findings, we propose a new integrated solution, named DAIL-SQL, which refreshes the Spider leaderboard with 86.6% execution accuracy and sets a new bar. To explore the potential of open-source LLM, we investigate them in various scenarios, and further enhance their performance with supervised fine-tuning. Our explorations highlight open-source LLMs' potential in Text-to-SQL, as well as the advantages and disadvantages of the supervised fine-tuning. Additionally, towards an efficient and economic LLM-based Text-to-SQL solution, we emphasize the token efficiency in prompt engineering and compare the prior studies under this metric. We hope that our work provides a deeper understanding of Text-to-SQL with LLMs, and inspires further investigations and broad applications.

In recent years, more attention is attracted by learning-based approaches in the field of video compression. Recent methods of this kind normally consist of three major components: intra-frame network, motion prediction network, and residual network, among which the motion prediction part is particularly critical for video compression. Benefiting from the optical flow which enables dense motion prediction, recent methods have shown competitive performance compared with traditional codecs. However, problems such as tail shadow and background distortion in the predicted frame remain unsolved. To tackle these problems, JMPNet is introduced in this paper to provide more accurate motion information by using both optical flow and dynamic local filter as well as an attention map to further fuse these motion information in a smarter way. Experimental results show that the proposed method surpasses state-of-the-art (SOTA) rate-distortion (RD) performance in the most data-sets.

Compression standards have been used to reduce the cost of image storage and transmission for decades. In recent years, learned image compression methods have been proposed and achieved compelling performance to the traditional standards. However, in these methods, a set of different networks are used for various compression rates, resulting in a high cost in model storage and training. Although some variable-rate approaches have been proposed to reduce the cost by using a single network, most of them brought some performance degradation when applying fine rate control. To enable variable-rate control without sacrificing the performance, we propose an efficient Interpolation Variable-Rate (IVR) network, by introducing a handy Interpolation Channel Attention (InterpCA) module in the compression network. With the use of two hyperparameters for rate control and linear interpolation, the InterpCA achieves a fine PSNR interval of 0.001 dB and a fine rate interval of 0.0001 Bits-Per-Pixel (BPP) with 9000 rates in the IVR network. Experimental results demonstrate that the IVR network is the first variable-rate learned method that outperforms VTM 9.0 (intra) in PSNR and Multiscale Structural Similarity (MS-SSIM).

Joint pushing and caching are commonly considered an effective way to adapt to tidal effects in networks. However, the problem of how to precisely predict users' future requests and push or cache the proper content remains to be solved. In this paper, we investigate a joint pushing and caching policy in a general mobile edge computing (MEC) network with multiuser and multicast data. We formulate the joint pushing and caching problem as an infinite-horizon average-cost Markov decision process (MDP). Our aim is not only to maximize bandwidth utilization but also to decrease the total quantity of data transmitted. Then, a joint pushing and caching policy based on hierarchical reinforcement learning (HRL) is proposed, which considers both long-term file popularity and short-term temporal correlations of user requests to fully utilize bandwidth. To address the curse of dimensionality, we apply a divide-and-conquer strategy to decompose the joint base station and user cache optimization problem into two subproblems: the user cache optimization subproblem and the base station cache optimization subproblem. We apply value function approximation Q-learning and a deep Q-network (DQN) to solve these two subproblems. Furthermore, we provide some insights into the design of deep reinforcement learning in network caching. The simulation results show that the proposed policy can learn content popularity very well and predict users' future demands precisely. Our approach outperforms existing schemes on various parameters including the base station cache size, the number of users and the total number of files in multiple scenarios.

Face recognition achieves great success thanks to the emergence of deep learning. However, many contemporary face recognition models still have limited invariance to strong intra-personal variations such as large pose changes. Face normalization provides an effective and cheap way to distil face identity and dispel face variances for recognition. We focus on face generation in the wild with unpaired data. To this end, we propose a Face Normalization Model (FNM) to generate a frontal, neutral expression, photorealistic face image for face recognition. FNM is a well-designed Generative Adversarial Network (GAN) with three distinct novelties. First, a face expert network is introduced to construct generator and provide the ability of retaining face identity. Second, with the reconstruction of normal face, pixel-wise loss is applied to stabilize optimization process. Third, we present a series of face attention discriminators to refine local textures. FNM could recover canonical-view, expression-free image and directly improve the performance of face recognition model. Extensive qualitative and quantitative experiments on both controlled and in-the-wild databases demonstrate the superiority of our face normalization method.

Pose and illumination variations are considered as two main challenges that face recognition system encounters. Most existing methods perform face normalization, aiming at untangling identity representation from these variations to improve recognition accuracy. Taking into account face variation representations, this paper proposes Task Specific Networks for the two representations with two novelties. First, we rotate and normalize face image to multi-pose view for one subtask, and learn face variation representations for another. Second, we learn face variation representations in an unsupervised way, which is more robust and more universal. We couple these two representations in the part of reconstructing the original face, where the two representations effect and restrict each other. Extensive experiments demonstrate the superiority of our method in both learning representations and rotating non-frontal face image.

How to quickly and consistently update a network is among the most fundamental and common challenges in software defined networking (SDN) systems. Current approaches heavily rely on the (logically) centralized controller to initiate and orchestrate the network updates, resulting in long latency of update completion. In this paper, we present DDP, a system for fast, distributed network updates while preserving various consistency properties. The key technique in DDP is a novel primitive named datapath operation container (DOC), where each DOC is encoded with an individual operation and its dependency logic. DDP adopts the simple, but powerful DOCs to configure the network, so that network updates can be triggered and executed at the data plane in a distributed and local manner. Novel algorithms are designed to compute and optimize the DOCs for consistent updates. We implement DDP to evaluate its performance in various update scenarios. Experimental results show that DDP significantly improves network update speed by up to 52.1% for the real-time updates initiated by the controller, and further improves the speed by 55.6-61.4% for the updates directly triggered at the data plane, such as failure recovery.