Search

The rise of multi-agent systems, especially the success of multi-agent reinforcement learning (MARL), is reshaping our future across diverse domains like autonomous vehicle networks. However, MARL still faces significant challenges, particularly in achieving zero-shot scalability, which allows trained MARL models to be directly applied to unseen tasks with varying numbers of agents. In addition, real-world multi-agent systems usually contain agents with different functions and strategies, while the existing scalable MARL methods only have limited heterogeneity. To address this, we propose a novel MARL framework named Scalable and Heterogeneous Proximal Policy Optimization (SHPPO), integrating heterogeneity into parameter-shared PPO-based MARL networks. we first leverage a latent network to adaptively learn strategy patterns for each agent. Second, we introduce a heterogeneous layer for decision-making, whose parameters are specifically generated by the learned latent variables. Our approach is scalable as all the parameters are shared except for the heterogeneous layer, and gains both inter-individual and temporal heterogeneity at the same time. We implement our approach based on the state-of-the-art backbone PPO-based algorithm as SHPPO, while our approach is agnostic to the backbone and can be seamlessly plugged into any parameter-shared MARL method. SHPPO exhibits superior performance over the baselines such as MAPPO and HAPPO in classic MARL environments like Starcraft Multi-Agent Challenge (SMAC) and Google Research Football (GRF), showcasing enhanced zero-shot scalability and offering insights into the learned latent representation's impact on team performance by visualization.

This paper introduces a novel approach for recovering sparse signals using sorted L1/L2 minimization. The proposed method assigns higher weights to indices with smaller absolute values and lower weights to larger values, effectively preserving the most significant contributions to the signal while promoting sparsity. We present models for both noise-free and noisy scenarios, and rigorously prove the existence of solutions for each case. To solve these models, we adopt a linearization approach inspired by the difference of convex functions algorithm. Our experimental results demonstrate the superiority of our method over state-of-the-art approaches in sparse signal recovery across various circumstances, particularly in support detection., Comment: 23 pages

We present a simple yet effective self-training approach, named as STAD, for low-resource relation extraction. The approach first classifies the auto-annotated instances into two groups: confident instances and uncertain instances, according to the probabilities predicted by a teacher model. In contrast to most previous studies, which mainly only use the confident instances for self-training, we make use of the uncertain instances. To this end, we propose a method to identify ambiguous but useful instances from the uncertain instances and then divide the relations into candidate-label set and negative-label set for each ambiguous instance. Next, we propose a set-negative training method on the negative-label sets for the ambiguous instances and a positive training method for the confident instances. Finally, a joint-training method is proposed to build the final relation extraction system on all data. Experimental results on two widely used datasets SemEval2010 Task-8 and Re-TACRED with low-resource settings demonstrate that this new self-training approach indeed achieves significant and consistent improvements when comparing to several competitive self-training systems. Code is publicly available at https://github.com/jjyunlp/STAD, Comment: Accepted by COLING'22

Electroencephalogram (EEG) recordings are often contaminated with artifacts. Various methods have been developed to eliminate or weaken the influence of artifacts. However, most of them rely on prior experience for analysis. Here, we propose an deep learning framework to separate neural signal and artifacts in the embedding space and reconstruct the denoised signal, which is called DeepSeparator. DeepSeparator employs an encoder to extract and amplify the features in the raw EEG, a module called decomposer to extract the trend, detect and suppress artifact and a decoder to reconstruct the denoised signal. Besides, DeepSeparator can extract the artifact, which largely increases the model interpretability. The proposed method is tested with a semi-synthetic EEG dataset and a real task-related EEG dataset, suggesting that DeepSeparator outperforms the conventional models in both EOG and EMG artifact removal. DeepSeparator can be extended to multi-channel EEG and data of any length. It may motivate future developments and application of deep learning-based EEG denoising. The code for DeepSeparator is available at https://github.com/ncclabsustech/DeepSeparator.

In recent years, distantly-supervised relation extraction has achieved a certain success by using deep neural networks. Distant Supervision (DS) can automatically generate large-scale annotated data by aligning entity pairs from Knowledge Bases (KB) to sentences. However, these DS-generated datasets inevitably have wrong labels that result in incorrect evaluation scores during testing, which may mislead the researchers. To solve this problem, we build a new dataset NYTH, where we use the DS-generated data as training data and hire annotators to label test data. Compared with the previous datasets, NYT-H has a much larger test set and then we can perform more accurate and consistent evaluation. Finally, we present the experimental results of several widely used systems on NYT-H. The experimental results show that the ranking lists of the comparison systems on the DS-labelled test data and human-annotated test data are different. This indicates that our human-annotated data is necessary for evaluation of distantly-supervised relation extraction., Comment: This paper has been accepted for publication in COLING2020

Mobile edge computing (MEC) provides users with a high quality experience (QoE) by placing servers with rich services close to the end users. Compared with local computing, MEC can contribute to energy saving, but results in increased communication latency. In this paper, we jointly optimize task offloading and resource allocation to minimize the energy consumption in an orthogonal frequency division multiple access (OFDMA)-based MEC networks, where the time-sensitive tasks can be processed at both local users and MEC server via partial offloading. Since the optimization variables of the problem are strongly coupled, we first decompose the original problem into two subproblems named as offloading selection (PO), and subcarriers and computing resource allocation (PS), and then propose an iterative algorithm to deal with them in a sequence. To be specific, we derive the closed-form solution for PO, and deal with PS by an alternating way in the dual domain due to its NP-hardness. Simulation results demonstrate, Comment: IEEE GLOBECOM 2020. arXiv admin note: substantial text overlap with arXiv:2003.12719

For active distribution networks (ADNs) integrated with massive inverter-based energy resources, it is impractical to maintain the accurate model and deploy measurements at all nodes due to the large-scale of ADNs. Thus, current models of ADNs are usually involving significant errors or even unknown. Moreover, ADNs are usually partially observable since only a few measurements are available at pilot nodes or nodes with significant users. To provide a practical Volt/Var control (VVC) strategy for such networks, a data-driven VVC method is proposed in this paper. Firstly, the system response policy, approximating the relationship between the control variables and states of monitoring nodes, is estimated by a recursive regression closed-form solution. Then, based on real-time measurements and the newly updated system response policy, a VVC strategy with convergence guarantee is realized. Since the recursive regression solution is embedded in the control stage, a data-driven closed-loop VVC framework is established. The effectiveness of the proposed method is validated in an unbalanced distribution system considering nonlinear loads where not only the rapid and self-adaptive voltage regulation is realized but also system-wide optimization is achieved., Comment: accepted by CSEE Journal of Power and Energy Systems and published in June 2021

Perfect optical vortices (POVs) provide an enabling solution to address the predicament induced by the strong dependence of classical optical vortices on theirs carried topological charges. Here, a type of generalized POVs with controllable impulse ring profile was proposed and demonstrated. Especially, a type of "absolute" dark POVs surrounded by two bright ringlobes in each side was presented, which provides a perfect annular potential well along those dark impulse rings for trapping steadily low-index particles, cells, or quantum gas, etc. In further, several POVs with different impulse ring profiles, including conventional POVs with bright rings, dark POVs mentioned above, and also POVs with controllable impulse ring profile, were demonstrated. This work opens up new possibilities to reshape arbitrarily the impulse ring profile for perfect vortices, and this type of novel POVs will enrich functions of optical vortices and it should be of high interest for its potential applications in optical manipulation, both quantum and classical optical communications, enhanced optical imaging, and also novel structured pumping lasers, etc.., Comment: There was a serious error in figure S1 and it has been corrected in the second version. Also, figure 4 was improved. I am the first time to use arXiv to upload the preprint version and I thought the replacement would 'replace' completely the first version. Afer replacement, I found the incorrect first version was still avaible for public. I am sorry for this trouble. Thanks