Data compression techniques significantly improve storage efficiency, bandwidth utilization, and energy efficiency, yet they introduce challenges for the rapid browsing and retrieval of valuable information within compressed data. Existing approaches achieve high-speed, lossless matching by exploiting the context-free property of automata. However, they are constrained by the recursive reference structures in compressed data, which necessitate state copying to ensure matching safety.

The rapid development of Web3.0 infrastructure has provided new approaches for building wisdom healthcare communities. However, existing solutions of federated learning struggle to handle large volumes of multi-modal privacy data across healthcare organizations. This paper proposes a blockchain-based verification mechanism for the entire process of multi-modal federated learning. For the initial federated state among multiple organizations, feature extraction, cosine similarity, and oracles are used to ensure privacy and security, while establishing a traceable initial state. Then, we perform encrypted verification for each iteration, utilize similarity matrices to realize mutual verification between clients, and validate the effectiveness of dynamically expanding multi-modal datasets from different organizations. This reduces on-chain storage and computational costs while protecting the privacy of organizations. Experimental results demonstrate that this mechanism effectively supports the needs of communities. Compared with state-of-the-art works, our method saves up to 70 % of on-chain storage and 54 % of computational costs, and enables the performance of medical federated learning training to differ from conventional models by no more than 1.89 %.

Wireless neural receivers provide a promising alternative to conventional receivers. To perform well in different channel environments, online adaption is required. However, during this process, performance remains low. Thus, an approach called federated collaborative learning with pruned-data aggregation (FedPDA) is developed to reduce online-adaptation frequency. The basic idea is that, upon online adaptation, mobile terminals further update their neural receivers collaboratively via federated learning. To reduce memory consumption, neural receivers follow a main-side network architecture where only the side network needs retraining during collaborative learning. To avoid catastrophic forgetting during continual learning, local data on terminals are pruned, with only a small percent sent to the base station. With such data, the base station also trains a neural receiver before conducting model aggregation. FedPDA is distinct with several features: 1) small memory footprint and no storage burden on terminals; 2) no catastrophic forgetting issue; 3) low communication cost. Performance results show that FedPDA reduces online adaptation by more than 90% and memory footprint by 70%. It achieves comparable performance as centralized schemes, but reducing communication cost by 78%. Compared to vanilla federated learning, FedPDA resolves the catastrophic forgetting issue without storage burden, and also reduces the communication cost by 50%.

The next generation of mobile networks are embracing disaggregation, reflected by the industry trend towards Open RAN. Private 5G networks are viewed as particularly suitable contenders for adopting Open RAN, owing to their setting, high degree of control, and opportunity for innovation. Motivated by this, we have recently deployed the first of its kind campus-wide, O-RAN-compliant private 5G testbed across the central campus of the University of Edinburgh. We first present the rationale behind our testbed along with an overview of its make-up. Then, we outline our plan to showcase the coverage, flexibility, and the operational view of the testbed from both network side and user perspectives.

An accurate prediction of energy production and consumption is a prerequisite for realizing reasonable energy scheduling in process industry. However, under the conditions of production-energy coupling, the energy operating status is highly related to production rhythm. Many prediction methods are unable to consider the impact of multi-production process correlation and planning constraints on energy data fluctuations. To tackle this problem, an industrial energy prediction method integrating plan and multi-dimensional data correlation is proposed. A data augmentation method based on wavelet matching is developed to extract specific features of the energy data and obtain augmented samples. To capture the alternating operation characteristics of different production processes, a contrastive learning (CL) method with probability jumping is developed that takes the process uncertainty into consideration. On this basis, the planning information is represented by a novel form of partial differential equations (PDEs), so that the global production information can be embedded as a priori knowledge within a physics-informed neural network (PINN) to achieve dynamic energy prediction. In order to validate the effectiveness of the proposed method, experiments are conducted using energy data from a steel company and compared with a variety of state-of-the-art methods. The results verify that the proposed method achieves better prediction results in complex industrial scenarios containing process coupling and planning constraints.

Backhaul mesh networks are critical for ensuring coverage and connectivity of high-frequency 6G networks. To maintain high throughput, its architecture needs to be augmented by adding relays. However, how to place relays at appropriate sites poses two challenges: 1) there lacks a theory to capture the relationship between a certain change of network architecture and its throughput gain; 2) selecting the best sites for relays is a complicated combinatorial problem. To tackle the first challenge, this paper first establishes a clique-based bottleneck theory, through which a clique-based bottleneck structure of a given network architecture is constructed to determine the network throughput. Based on this bottleneck structure, clique gradients are then computed to quantify the impact of each clique on the overall network throughput. With the clique-based bottleneck theory, the second challenge is resolved by embedding clique gradients into a deep reinforcement learning (DRL) scheme. Specifically, the DRL actions are masked such that only the relay sites that match the highest clique gradients are selected. This DRL-based relay placement (DeepRP) scheme is evaluated via extensive simulations, and performance results show that it can boost network throughput by more than 50%, which is$\text{10.4} \!-\! \text{32.1}\% $higher than those of baseline schemes.

Terahertz (THz) mesh networking is envisioned as a promising technology for 6G networks, with network capacity as one of the most critical performance metrics. To boost the network capacity of a THz mesh network, link resource planning is conducted, which poses two challenges. First, the relationship between link resources and network capacity must be captured quantitatively considering the peculiarities of THz mesh networking. Second, multi-dimensional resources including subarrays, power, and subbands need to be determined for link resource planning. To address the first challenge, a bottleneck structure is constructed by adapting the quantitative theory of bottleneck structures in the recent work [1] for THz mesh networks, such that the relationship between network capacity and a certain link resource planning result is determined. Furthermore, bottleneck gradients are computed based on the constructed bottleneck structure. Given the derived relationship and bottleneck gradients, a heuristic link resource planning algorithm is designed to allocate multi-dimensional resources, thus resolving the second challenge. Performance results show that the heuristic resource planning algorithm can boost the network capacity by 20.3% - 41.8% for various topologies.

Personalization plays a critical role in helping customers discover the products and contents they prefer for e-commerce stores.Personalized recommendations differ in contents, target customers, and UI. However, they require a common core capability - the ability to deeply understand customers’ preferences and shopping intents. In this paper, we introduce the MCM (Multi-task pre-trained Customer Model), a large pre-trained BERT-based multi-task customer model with 10 million trainable parameters for e-commerce stores. This model aims to empower all personalization projects by providing commonly used preference scores for recommendations, customer embeddings for transfer learning, and a pre-trained model for fine-tuning. In this work, we improve the SOTA BERT4Rec framework to handle heterogeneous customer signals and multi-task training as well as innovate new data augmentation method that is suitable for recommendation task. Experimental results show that MCM outperforms the original BERT4Rec by 17% on on NDCG@10 of next action prediction tasks. Additionally, we demonstrate that the model can be easily fine-tuned to assist a specific recommendation task. For instance, after fine-tuning MCM for an incentive based recommendation project, performance improves by 60% on the conversion prediction task and 25% on the click-through prediction task compared to a baseline tree-based GBDT model.

Considering network slicing in a cellular network, one of the most intriguing tasks is slice enforcement over air interfaces across multiple cells. The challenges lie in several aspects. First, resources allocated to different slices must achieve soft isolation at the link level. Second, users’ diverse QoS requirements must be satisfied even when communication links experience fading and interference. Third, long-term slicing policies must be conformed, no matter how unbalanced they are. To address these challenges, link-level slice enforcement is first formulated as a resource allocation problem that minimizes radio resource consumption while ensuring link-level soft slice isolation, guaranteeing users’ diverse QoS requirements, and conforming to slicing policies. Next, this problem is tackled via a deep reinforcement learning (DRL) based approach, through which LinkSlice is designed as an iterative two-stage algorithm. The first stage determines transmission rates for each link based on DRL. It is embedded with a graph neural network (GNN) to characterize link interference. Based on the transmission rates from the first stage, the second stage allocates resources to each slice. Performance results show that LinkSlice converges quickly to a near-optimal solution. It gracefully tackles the three challenges of link-level slice enforcement while further improving throughput by 18.5%.

Cold start problem is one of the most challenging and long-standing problems in recommender systems, and cross-domain recommendation (CDR) methods are effective for tackling it. Most cold-start related CDR methods require training a mapping function between high-dimensional embedding space using overlapping user data. However, the overlapping data is scarce in many recommendation tasks, which makes it difficult to train the mapping function. In this paper, we propose a new approach for CDR, which aims to alleviate the training difficulty. The proposed method can be viewed as a special parameterization of the mapping function without hurting expressiveness, which makes use of non-overlapping user data and leads to effective optimization. Extensive experiments on two real-world CDR tasks are performed to evaluate the proposed method. In the case that there are few overlapping data, the proposed method outperforms the existed state-of-the-art method by 14% (relative improvement).

In article recommendation, the amount of time user spends on viewing articles, dwell time, is an important metric to measure the post-click engagement of user on content and has been widely used as a proxy to user satisfaction, complementing the click feedback. Recently, the sequential pattern of impression-click-read has become one of the most popular type of article recommendation service in real world, where users are presented with a list of titles at first, then get interested in one and click in for reading. Predicting dwell time in such service is conditioned on the click, since the user reads the article only after he clicks the corresponding title. We argue that conventional models for dwell time prediction, which mainly focus on the relevance between the content and the general preference of user, are not well-designed for such service. There is a natural assumption in recommendation system that the click indicates user's getting attracted by the item. Therefore, in the pattern of impression-click-read, the user might get interested and curious on some other concepts different from his general preference while reading, due to the attraction of the title. Conventional models tend to ignore the gap between such temporary interest and the general preference of user in the reading behavior, which fails to use the pattern of impression-click-read and the assumption of the click very well. In this work, we propose a framework, Click-guide Network (CGN) for dwell time prediction, which makes good use of the sequential pattern and the assumption to model the ”guidance” of the click on user preference. CGN is a joint learner for dwell time and click through rate (CTR). We introduce the CTR task as an auxiliary task to help us better learn the preference of user and the representation of title. Besides, we propose the Guider to capture the user's temporary interest raised by the title. We collect the data from WeChat, a widely-used mobile app in China, for experiments. The results demonstrate the advantages of CGN over several competitive baselines on dwell time prediction, while our case studies show how the Guider effectively capture the temporary interest of user.

In article recommendation, the dwell time is an important metric to measure user engagement on content and has been widely used as a proxy for user satisfaction. Therefore, predicting the dwell time is very helpful for making better recommendations and improving user experience. Modeling the interaction between user and content is the key for dwell time prediction. However, conventional methods usually model the content with document-level representation in a non-personalized way, which ignores the natural reading process of the reader and the reader attraction in sub-document level, this might lead to a bias for analyzing the user reading behavior. Since the attraction level of different parts is different for the user, the user attention changes dynamically while reading. The former content affects the reading for the latter content via the change of attraction level. Therefore, considering the attraction level of each part of the article, i.e., attraction distribution, is quite necessary for content modeling. In this paper, we propose the Sequential Attentive Network (SAN) for dwell time prediction, which effectively models the attraction distribution of the article reading process. We collect the data from WeChat, a widely-used mobile app in China, for experiments. The results demonstrate the advantages of our model over several competitive baselines on dwell time prediction.