Machine learning has been adapted to help solve NP-hard combinatorial optimization problems. One prevalent way is learning to construct solutions by deep neural networks, which has been receiving more and more attention due to the high efficiency and less requirement for expert knowledge. However, many neural construction methods for Vehicle Routing Problems (VRPs) focus on synthetic problem instances with limited scales and specified node distributions, leading to poor performance on real-world problems which usually involve large scales together with complex and unknown node distributions. To make neural VRP solvers more practical in real-world scenarios, we design an auxiliary policy that learns from the local transferable topological features, named local policy, and integrate it with a typical constructive policy (which learns from the global information of VRP instances) to form an ensemble policy. With joint training, the aggregated policies perform cooperatively and complementarily to boost generalization. The experimental results on two well-known benchmarks, TSPLIB and CVRPLIB, of travelling salesman problem and capacitated VRP show that the ensemble policy consistently achieves better generalization than state-of-the-art construction methods and even works well on real-world problems with several thousand nodes.
Federated learning (FL) can lead to significant communication overhead and reliance on a central server. To address these challenges, decentralized federated learning (DFL) has been proposed as a more resilient framework. DFL involves parameter exchange between devices through a wireless network. This study analyzes the performance of resource-constrained DFL using different communication schemes (digital and analog) over wireless networks to optimize communication efficiency. Specifically, we provide convergence bounds for both digital and analog transmission approaches, enabling analysis of the model performance trained on DFL. Furthermore, for digital transmission, we investigate and analyze resource allocation between computation and communication and convergence rates, obtaining its communication complexity and the minimum probability of correction communication required for convergence guarantee. For analog transmission, we discuss the impact of channel fading and noise on the model performance and the maximum errors accumulation with convergence guarantee over fading channels. Finally, we conduct numerical simulations to evaluate the performance and convergence rate of convolutional neural networks (CNNs) and Vision Transformer (ViT) trained in the DFL framework on fashion-MNIST and CIFAR-10 datasets. Our simulation results validate our analysis and discussion, revealing how to improve performance by optimizing system parameters under different communication conditions.
Graph Neural Network (GNN)-based models have become the mainstream approach for recommender systems. Despite the effectiveness, they are still suffering from the cold-start problem, i.e., recommend for few-interaction items. Existing GNN-based recommendation models to address the cold-start problem mainly focus on utilizing auxiliary features of users and items, leaving the user-item interactions under-utilized. However, embeddings distributions of cold and warm items are still largely different, since cold items' embeddings are learned from lower-popularity interactions, while warm items' embeddings are from higher-popularity interactions. Thus, there is a seesaw phenomenon, where the recommendation performance for the cold and warm items cannot be improved simultaneously. To this end, we proposed a Uncertainty-aware Consistency learning framework for Cold-start item recommendation (shorten as UCC) solely based on user-item interactions. Under this framework, we train the teacher model (generator) and student model (recommender) with consistency learning, to ensure the cold items with additionally generated low-uncertainty interactions can have similar distribution with the warm items. Therefore, the proposed framework improves the recommendation of cold and warm items at the same time, without hurting any one of them. Extensive experiments on benchmark datasets demonstrate that our proposed method significantly outperforms state-of-the-art methods on both warm and cold items, with an average performance improvement of 27.6%.
We present TransNormerLLM, the first linear attention-based Large Language Model (LLM) that outperforms conventional softmax attention-based models in terms of both accuracy and efficiency. TransNormerLLM evolves from the previous linear attention architecture TransNormer by making advanced modifications that include positional embedding, linear attention acceleration, gating mechanism, tensor normalization, inference acceleration and stabilization. Specifically, we use LRPE together with an exponential decay to avoid attention dilution issues while allowing the model to retain global interactions between tokens. Additionally, we propose Lightning Attention, a cutting-edge technique that accelerates linear attention by more than twice in runtime and reduces memory usage by a remarkable four times. To further enhance the performance of TransNormer, we leverage a gating mechanism to smooth training and a new tensor normalization scheme to accelerate the model, resulting in an impressive acceleration of over 20%. Furthermore, we have developed a robust inference algorithm that ensures numerical stability and consistent inference speed, regardless of the sequence length, showcasing superior efficiency during both training and inference stages. Scalability is at the heart of our model's design, enabling seamless deployment on large-scale clusters and facilitating expansion to even more extensive models, all while maintaining outstanding performance metrics. Rigorous validation of our model design is achieved through a series of comprehensive experiments on our self-collected corpus, boasting a size exceeding 6TB and containing over 2 trillion tokens. To ensure data quality and relevance, we implement a new self-cleaning strategy to filter our collected data. Our pre-trained models will be released to foster community advancements in efficient LLMs.
Social media streams contain large and diverse amount of information, ranging from daily-life stories to the latest global and local events and news. Twitter, especially, allows a fast spread of events happening real time, and enables individuals and organizations to stay informed of the events happening now. Event detection from social media data poses different challenges from traditional text and is a research area that has attracted much attention in recent years. In this paper, we survey a wide range of event detection methods for Twitter data stream, helping readers understand the recent development in this area. We present the datasets available to the public. Furthermore, a few research opportunities
As machine learning and deep learning models become ubiquitous, it is inevitable that there will be attempts to exploit such models in various attack scenarios. For example, in a steganographic-based attack, information could be hidden in a learning model, which might then be used to distribute malware, or for other malicious purposes. In this research, we consider the steganographic capacity of several learning models. Specifically, we train a Multilayer Perceptron (MLP), Convolutional Neural Network (CNN), and Transformer model on a challenging malware classification problem. For each of the resulting models, we determine the number of low-order bits of the trained parameters that can be altered without significantly affecting the performance of the model. We find that the steganographic capacity of the learning models tested is surprisingly high, and that in each case, there is a clear threshold after which model performance rapidly degrades.
Cognitive radio has been proposed to alleviate the scarcity of available spectrum. However, sensing performance is quite poor due to the low sensing signal-to-noise ratio. Fortunately, reconfigurable intelligent surface (RIS)-aided spectrum sensing can effectively tackle the above challenge due to its high array gain. Nevertheless, the traditional passive RIS suffers from the ``double fading'' effect, which severely restricts the performance of passive RIS-aided spectrum sensing. To this end, we introduce the active RIS into spectrum sensing and respectively formulate two optimization problems for the passive RIS and the active RIS to maximize the detection probability. In light of the intractability of the formulated problems, we develop a one-stage optimization algorithm with inner approximation and a two-stage optimization algorithm with a bisection method to obtain sub-optimal solutions, and apply the Rayleigh quotient to obtain the upper and lower bounds of the detection probability. Furthermore, in order to gain more insight into the impact of the RIS on spectrum sensing, we respectively investigate the number configuration for passive RIS and active RIS and analyze how many reflecting elements are needed to achieve the detection probability close to 1. Simulation results verify the effectiveness of the proposed algorithms.
This letter proposes a new user cooperative offloading protocol called user reciprocity in backscatter communication (BackCom)-aided mobile edge computing systems with efficient computation, whose quintessence is that each user can switch alternately between the active or the BackCom mode in different slots, and one user works in the active mode and the other user works in the BackCom mode in each time slot. In particular, the user in the BackCom mode can always use the signal transmitted by the user in the active mode for more data transmission in a spectrum-sharing manner. To evaluate the proposed protocol, a computation efficiency (CE) maximization-based optimization problem is formulated by jointly power control, time scheduling, reflection coefficient adjustment, and computing frequency allocation, while satisfying various physical constraints on the maximum energy budget, the computing frequency threshold, the minimum computed bits, and harvested energy threshold. To solve this non-convex problem, Dinkelbach's method and quadratic transform are first employed to transform the complex fractional forms into linear ones. Then, an iterative algorithm is designed by decomposing the resulting problem to obtain the suboptimal solution. The closed-form solutions for the transmit power, the RC, and the local computing frequency are provided for more insights. Besides, the analytical performance gain with the reciprocal mode is also derived. Simulation results demonstrate that the proposed scheme outperforms benchmark schemes regarding the CE.
Sequence modeling has important applications in natural language processing and computer vision. Recently, the transformer-based models have shown strong performance on various sequence modeling tasks, which rely on attention to capture pairwise token relations, and position embedding to inject positional information. While showing good performance, the transformer models are inefficient to scale to long input sequences, mainly due to the quadratic space-time complexity of attention. To overcome this inefficiency, we propose to model sequences with a relative position encoded Toeplitz matrix and use a Toeplitz matrix-vector production trick to reduce the space-time complexity of the sequence modeling to log linear. A lightweight sub-network called relative position encoder is proposed to generate relative position coefficients with a fixed budget of parameters, enabling the proposed Toeplitz neural network to deal with varying sequence lengths. In addition, despite being trained on 512-token sequences, our model can extrapolate input sequence length up to 14K tokens in inference with consistent performance. Extensive experiments on autoregressive and bidirectional language modeling, image modeling, and the challenging Long-Range Arena benchmark show that our method achieves better performance than its competitors in most downstream tasks while being significantly faster. The code is available at https://github.com/OpenNLPLab/Tnn.
Reconfigurable intelligent surface (RIS) has been regarded as a promising technique due to its high array gain and low power. However, the traditional passive RIS suffers from the ``double fading'' effect, which has restricted the performance of passive RIS-aided communications. Fortunately, active RIS can alleviate this problem since it can adjust the phase shift and amplify the received signal simultaneously. Nevertheless, a high beamforming gain often requires a number of reflecting elements, which leads to non-negligible power consumption, especially for the active RIS. Thus, one challenge is how to improve the scalability of the RIS and the energy efficiency. Different from the existing works where all reflecting elements are activated, we propose a novel element on-off mechanism where reflecting elements can be flexibly activated and deactivated. Two different optimization problems for passive RIS and active RIS are formulated by maximizing the total energy efficiency. We develop two different alternating optimization-based iterative algorithms to obtain sub-optimal solutions. Furthermore, we consider special cases involving rate maximization problems for given the same total power budget, and respectively analyze the number configuration for passive RIS and active RIS. Simulation results verify that reflecting elements under the proposed algorithms can be flexibly activated and deactivated.