Many deep learning architectures have been employed to model the semantic compositionality for text sequences, requiring a huge amount of supervised data for parameters training, making it unfeasible in situations where numerous annotated samples are not available or even do not exist. Different from data-hungry deep models, lightweight word embedding-based models could represent text sequences in a plug-and-play way due to their parameter-free property. In this paper, a modified hierarchical pooling strategy over pre-trained word embeddings is proposed for text classification in a few-shot transfer learning way. The model leverages and transfers knowledge obtained from some source domains to recognize and classify the unseen text sequences with just a handful of support examples in the target problem domain. The extensive experiments on five datasets including both English and Chinese text demonstrate that the simple word embedding-based models (SWEMs) with parameter-free pooling operations are able to abstract and represent the semantic text. The proposed modified hierarchical pooling method exhibits significant classification performance in the few-shot transfer learning tasks compared with other alternative methods.

Unmanned aerial vehicles (UAVs) are commonly employed in pursuit and rescue missions, where the target’s trajectory is unknown. Traditional methods, such as evolutionary algorithms and ant colony optimization, can generate a search route in a given scenario. However, when the scene changes, the solution needs to be recalculated. In contrast, more advanced deep reinforcement learning methods can train an agent that can be directly applied to a similar task without recalculation. Nevertheless, there are several challenges when the agent learns how to search for unknown dynamic targets. In this search task, the rewards are random and sparse, which makes learning difficult. In addition, because of the need for the agent to adapt to various scenario settings, interactions required between the agent and the environment are more comparable to typical reinforcement learning tasks. These challenges increase the difficulty of training agents. To address these issues, we propose the OC-MAPPO method, which combines optimal control (OC) and Multi-Agent Proximal Policy Optimization (MAPPO) with GPU parallelization. The optimal control model provides the agent with continuous and stable rewards. Through parallelized models, the agent can interact with the environment and collect data more rapidly. Experimental results demonstrate that the proposed method can help the agent learn faster, and the algorithm demonstrated a 26.97% increase in the success rate compared to genetic algorithms.

Unmanned aerial vehicles (UAVs) are becoming more and more widely used in battlefield reconnaissance and target strikes because of their high cost-effectiveness, but task planning for large-scale UAV swarms is a problem that needs to be solved. To solve the high-risk problem caused by incomplete information for the combat area and the potential coordination between targets when a heterogeneous UAV swarm performs reconnaissance and strike missions, this paper proposes a distributed task-allocation algorithm. The method prioritizes tasks by evaluating the swarm’s capability superiority to tasks to reduce the search space, uses the time coordination mechanism and deterrent maneuver strategy to reduce the risk of reconnaissance missions, and uses the distributed negotiation mechanism to allocate reconnaissance tasks and coordinated strike tasks. The simulation results under the distributed framework verify the effectiveness of the distributed negotiation mechanism, and the comparative experiments under different strategies show that the time coordination mechanism and the deterrent maneuver strategy can effectively reduce the mission risk when the target is unknown. The comparison with the centralized global optimization algorithm verifies the efficiency and effectiveness of the proposed method when applied to large-scale UAV swarms. Since the distributed negotiation task-allocation architecture avoids dependence on the highly reliable network and the central node, it can further improve the reliability and scalability of the swarm, and make it applicable to more complex combat environments.

Identifying the most important nodes is significant for investigating the robustness and vulnerability of complex network. A lot of methods based on network structure have been proposed, such as degree, K-shell and betweenness, etc. In order to identify the important nodes in a more reasonable way, both the network topologies and the characteristics of nodes should be taken into account. Even at the same location, the nodes with different characteristics have different importance. The topological structures and the characteristics of the nodes are considered in the complex network dynamics model. However, such methods are rarely explored and their applications are restricted. In order to identify the important nodes in undirected weighted networks, in this paper we propose a method based on dynamics model. Firstly, we introduce a way to construct the corresponding dynamics model for any undirected weighted network, and the constructed model can be flexibly adjusted according to the actual situation. It is proved that the constructed model is globally asymptotic stable. To measure the changes of the dynamic model state, the mean deviation and the variance are presented, which are the criteria to evaluate the importance of the nodes. Finally, disturbance test and destructive test are proposed for identifying the most important nodes. Each node is tested in turn, and then the important nodes are identified. If the tested node can recover from the damaged state, the disturbance test is used. If the tested node is destroyed completely, the destructive test is used. The method proposed in this paper is based on the dynamics model. The node importance is influenced by the network topologies and the characteristics of nodes in these two methods. In addition, the disturbance test and destructive test are used in different situations, forming a complementary advantage. So the method can be used to analyze the node importance in a more comprehensive way. Experiments are performed on the advanced research project agency networks, the undirected networks with symmetric structures, the social network, the Dobbs-Watts-Sabel networks and the Barrat-Barthelemy-Vespignani networks. If the nodes in the network have the same dynamic model, the network is considered to be the homogeneous network; otherwise, the network is heterogeneous network. And experiments can be divided into four categories, namely, the disturbance test, the destructive test on the homogeneous network, the disturbance test and the destructive test on the heterogeneous network. The experimental results show that the methods proposed in this paper are effective and credible.