The latent factor analysis (LFA) model has been widely used to uncover latent relationships from high-dimensional sparse (HiDS) matrices. However, the performance of LFA depends largely on the hyper-parameter value used in the model training. Traditional hyper-parameter tuning methods such as grid search suffer from inefficiency and inaccuracy. In recent years, the particle swarm optimization (PSO) algorithm offers an intelligent approach to adaptively adjust the hyper-parameter of LFA. However, the global optimal solution of the hyper-parameter tuning problem is not fixed due to its dynamic decision space. Therefore, it is difficult for PSO to determine the best hyper-parameter for each training iteration. To address this problem, this paper proposes a novel hyper-parameter adaptive adjustment algorithm called dynamic stochastic reorientation PSO (DSR-PSO) that adapts to constantly changing decision spaces. By randomly adjusting the search directions of particles and perturbing the elite particles, the dynamic property of the DSR-PSO can be enhanced, so that the hyper-parameter can be adjusted in real time throughout the model training process. Furthermore, this paper proves the convergence of the DSR-PSO and gives its convergence condition by discussing the distribution of the characteristic roots. Finally, this paper proposes the DSR-PSO-based LFA (DPL) model by incorporating the DSR-PSO-based hyper-parameter adjustment into the LFA to promote its model training, and analyzes its complexity. Experimental results on benchmark datasets show that the proposed DPL surpasses state-of-the-art LFA models in terms of accuracy and efficiency.

The latent factor analysis (LFA) model is an effective tool for extracting valuable information from high-dimensional and sparse (HiDS) matrices. However, traditional LFA usually suffers from low accuracy due to the limitations of the stochastic gradient descent (SGD) algorithm used in its model training. First, the learning rate of SGD is adjusted manually, which greatly affects the training efficiency. A recent solution is adjusting this hyperparameter by the particle swarm optimization (PSO) algorithm. However, PSO cannot adapt well to the dynamic decision space of this problem due to its strong convergence. Second, SGD relies solely on the gradient information to perform the optimization, which decreases the training accuracy. To address the above two issues, this article proposes a novel LFA model called genetic algorithm-based two-step LFA (GA-TSLFA), which employs the GA to facilitate the model training. Compared to PSO, the GA has better flexibility, which can be employed to tune the hyperparameter of LFA in dynamic decision spaces and refine the model in high-dimensional and complex decision spaces by designing suitable evolutionary operators. The training of the proposed GA-TSLFA consists of two steps. In the first step, the model is pretrained by SGD whose learning rate is adaptively adjusted by a proposed GA. In the second step, the LF matrices generated by SGD are further refined using a proposed GA-based framework. This framework operates by optimizing a subset of partial vectors, which are selected through a dedicated strategy. In this way, the model's accuracy can be further enhanced. Empirical studies on benchmark datasets show that the GA-TSLFA surpasses state-of-the-art LFA models in prediction accuracy and has a competitive efficiency.

Human-robot interactive target-oriented grasping in unstructured environments, guided by natural language, is crucial for enabling intelligent robotic arms to perform tasks safely and efficiently. However, it remains a challenge for robotic arms to comprehend human instructions and execute corresponding grasping actions. In this paper, we propose QwenGrasp, a novel system that uses a large vision-language model to align workspace images with textual instructions. This alignment enables QwenGrasp to perform accurate 6DoF grasping on the specified target object. Additionally, we introduce Masked REGNet, which incorporates target-object location information into the network to generate precise grasp poses and ensure high grasp quality. Through extensive real-world experiments, QwenGrasp achieves over 90% success across six diverse instruction types. The results highlight QwenGrasp’s ability to understand human intent and execute precise grasping actions. Notably, it outperforms other target-oriented methods in both performance and instruction comprehension. Even when given vague descriptions, directional cues, or complex instructions, QwenGrasp reliably identifies and grasps the correct object. An ablation study further confirms the importance of each component, with all contributing significantly to robust and high-quality grasping.

The rapid expansion of location-based services has triggered the acquisition and analysis of various types of individual trajectory recordings. However, the sensitive nature of such kind of data inevitably leads to privacy constraints and regulations on its use and sharing. This paper addresses the problem in a distinctive perspective. Instead of blurring or modifying original trajectory samples, we aim to generate a completely synthetic dataset, whose samples are singularly different from the original ones, but whose collective sets share similar global characteristics and performances. We propose a generative deep learning solution for location-based trajectory formats, with the goal of producing realistic synthetic location sequences: the process relies on a generative adversarial network (GAN) framework, involving long short-term memory (LSTM) recurrent layers to capture trajectory characteristics, and neural embeddings to model mobility relations between places. We leverage multiple metrics to assess the realistic character of synthetic data and their similarity with the original source; moreover, we evaluate downstream performance differences with regard to the next place prediction problem. Tested on a real-world large-scale dataset of long-distance trips, and compared with baselines and traditional geomasking techniques, our approach presents better characteristics, providing novel insights into GeoAI solutions for human mobility analysis.

Abstract Recommender systems have been widely deployed in various real-world applications to help users identify content of interest from massive amounts of information. Traditional recommender systems work by collecting user-item interaction data in a cloud-based data center and training a centralized model to perform the recommendation service. However, such cloud-based recommender systems (CloudRSs) inevitably suffer from excessive resource consumption, response latency, as well as privacy and security risks concerning both data and models. Recently, driven by the advances in storage, communication, and computation capabilities of edge devices, there has been a shift of focus from CloudRSs to on-device recommender systems (DeviceRSs), which leverage the capabilities of edge devices to minimize centralized data storage requirements, reduce the response latency caused by communication overheads, and enhance user privacy and security by localizing data processing and model training. Despite the rapid rise of DeviceRSs, there is a clear absence of timely literature reviews that systematically introduce, categorize and contrast these methods. To bridge this gap, we aim to provide a comprehensive survey of DeviceRSs, covering three main aspects: (1) the deployment and inference of DeviceRSs, exploring how large recommendation models can be compressed and utilized within resource-constrained on-device environments; (2) the training and update of DeviceRSs, discussing how local data can be leveraged for model optimization on the device side; (3) the security and privacy of DeviceRSs, unveiling their potential vulnerability to malicious attacks and defensive strategies to safeguard these systems. Furthermore, we provide a fine-grained and systematic taxonomy of the methods involved in each aspect, followed by a discussion regarding challenges and future research directions. This is the first comprehensive survey on DeviceRSs that covers a spectrum of tasks to fit various needs. We believe this survey will help readers understand the current research status in this field, equip them with relevant technical foundations, and stimulate new research ideas for developing DeviceRSs.

Automated design of metaheuristic algorithms offers an attractive avenue to reduce human effort and gain enhanced performance beyond human intuition. Current automated methods design algorithms within a fixed structure and operate from scratch. This poses a clear gap toward fully discovering potentials over the metaheuristic family and fertilizing from prior design experience. To bridge the gap, this article proposes an autoregressive learning-based designer for automated design of metaheuristic algorithms. Our designer formulates metaheuristic algorithm design as a sequence generation task, and harnesses an autoregressive generative network to handle the task. This offers two advances. First, through autoregressive inference, the designer generates algorithms with diverse lengths and structures, enabling to fully discover potentials over the metaheuristic family. Second, prior design knowledge learned and accumulated in neurons of the designer can be retrieved for designing algorithms for future problems, paving the way to continual design of algorithms for open-ended problem solving. Extensive experiments on numeral benchmarks and real-world problems reveal that the proposed designer generates algorithms that outperform all human-created baselines on 24 out of 25 test problems. The generated algorithms display various structures and behaviors, reasonably fitting for different problem-solving contexts. Code is available athttps://github.com/auto4opt/ALDes.

Gradient inversion attacks (GIAs) pose significant challenges to the privacy-preserving paradigm of distributed learning. These attacks employ carefully designed strategies to reconstruct victim’s private training data from their shared gradients. However, existing work mainly focuses on attacks and defenses for image-modal data, while the study for text-modal data remains scarce. Furthermore, the performance of the limited attack researches on text-modal data is also unsatisfactory, which can be partially attributed to the finer granularity of text data compared to image. To bridge the existing research gap, we propose a high-fidelity attack method tailored for Transformer-based language models (LMs). In our method, we initially reconstruct the label space of the victim’s training data by leveraging the characteristics of the Transformer architecture. After that, we propose a shallow-to-deep paradigm to facilitate gradient matching, which can significantly improve the attack performance. Furthermore, we develop a weighted surrogate loss that resolves the consistent deviation issue present in current attack researches. A substantial number of experiments on Transformer-based LMs (e.g., Bert and GPT) demonstrate that our attack is competitive and significantly outperforms existing methods. In the final part of this paper, we investigate the influence of the inherent position embedding module within the Transformer architecture on attack performance, and based on the analysis results, we propose a countermeasure to alleviate part of the privacy leakage issue in distributed learning.

The application of deep learning techniques to medical problems has garnered widespread research interest in recent years, such as applying convolutional neural networks to medical image classification tasks. However, data in the medical field is often highly private, preventing different hospitals from sharing data to train an accurate model. Federated learning, as a privacy-preserving machine learning architecture, has shown promising performance in balancing data privacy and model utility by keeping private data on the client's side and using a central server to coordinate a set of clients for model training through aggregating their uploaded model parameters. Yet, this architecture heavily relies on a trusted third-party server, which is challenging to achieve in real life. Swarm learning, as a specialized decentralized federated learning architecture that does not require a central server, utilizes blockchain technology to enable direct parameter exchanges between clients. However, the mining of blocks requires significant computational resources, limiting its scalability. To address this issue, this paper integrates the brain storm optimization algorithm into the swarm learning framework, named BSO-SL. This approach clusters similar clients into different groups based on their model distributions. Additionally, leveraging the architecture of BSO, clients are given the probability to engage in collaborative learning both within their cluster and with clients outside their cluster, preventing the model from converging to local optima. The proposed method has been validated on a real-world diabetic retinopathy image classification dataset, and the experimental results demonstrate the effectiveness of the proposed approach.

While large language models (LLMs) excel in a simulated world of texts, they struggle to interact with the more realistic world without perceptions of other modalities such as visual or audio signals. Although vision-language models (VLMs) integrate LLM modules (1) aligned with static image features, and (2) may possess prior knowledge of world dynamics (as demonstrated in the text world), they have not been trained in an embodied visual world and thus cannot align with its dynamics. On the other hand, training an embodied agent in a noisy visual world without expert guidance is often chal-lenging and inefficient. In this paper, we train a VLM agent living in a visual world using an LLM agent excelling in a parallel text world. Specifically, we distill LLM's reflection outcomes (improved actions by analyzing mistakes) in a text world's tasks to finetune the VLM on the same tasks of the visual world, resulting in an Embodied Multi-Modal Agent (EMMA) quickly adapting to the visual world dy-namics. Such cross-modality imitation learning between the two parallel worlds is achieved by a novel DAgger-DPO algorithm, enabling EMMA to generalize to a broad scope of new tasks without any further guidance from the LLM expert. Extensive evaluations on the ALFWorld benchmark's diverse tasks highlight EMMA's superior performance to SOTA VLM-based agents, e.g., 20%-70% improvement in the success rate.

Despite advancements in Text-to-Video (T2V) generation, producing videos with realistic motion remains challenging. Current models often yield static or minimally dynamic outputs, failing to capture complex motions described by text. This issue stems from the internal biases in text encoding which overlooks motions, and inadequate conditioning mechanisms in T2V generation models. To address this, we propose a novel framework called DEcomposed MOtion (DEMO), which enhances motion synthesis in T2V generation by decomposing both text encoding and conditioning into content and motion components. Our method includes a content encoder for static elements and a motion encoder for temporal dynamics, alongside separate content and motion conditioning mechanisms. Crucially, we introduce text-motion and video-motion supervision to improve the model's understanding and generation of motion. Evaluations on benchmarks such as MSR-VTT, UCF-101, WebVid-10M, EvalCrafter, and VBench demonstrate DEMO's superior ability to produce videos with enhanced motion dynamics while maintaining high visual quality. Our approach significantly advances T2V generation by integrating comprehensive motion understanding directly from textual descriptions. Project page: https://PR-Ryan.github.io/DEMO-project/

To make room for privacy and efficiency, the deployment of many recommender systems is experiencing a shift from central servers to personal devices, where the federated recommender systems (FedRecs) and decentralized collaborative recommender systems (DecRecs) are arguably the two most representative paradigms. While both leverage knowledge (e.g., gradients) sharing to facilitate learning local models, FedRecs rely on a central server to coordinate the optimization process, yet in DecRecs, the knowledge sharing directly happens between clients. On the flip side, knowledge sharing also opens a backdoor for model poisoning attacks, where adversaries disguise themselves as benign clients and disseminate polluted knowledge to achieve malicious goals like promoting an item's exposure rate. Although research on such poisoning attacks provides valuable insights into finding security loopholes and corresponding countermeasures, existing attacks mostly focus on FedRecs, and are either inapplicable or ineffective for DecRecs. Compared with FedRecs where the tampered information can be universally distributed to all clients once uploaded to the cloud, each adversary in DecRecs can only communicate with neighbor clients of a small size, confining its impact to a limited range.

Federated recommender systems (FedRecs) have gained significant attention for their potential to protect user's privacy by keeping user privacy data locally and only communicating model parameters/gradients to the server. Nevertheless, the currently existing architecture of FedRecs assumes that all users have the same 0-privacy budget, i.e., they do not upload any data to the server, thus overlooking those users who are less concerned about privacy and are willing to upload data to get a better recommendation service. To bridge this gap, this paper explores a user-governed data contribution federated recommendation architecture where users are free to take control of whether they share data and the proportion of data they share to the server. To this end, this paper presents a cloud-device collaborative graph neural network federated recommendation model, named CDCGNNFed. It trains user-centric ego graphs locally, and high-order graphs based on user-shared data in the server in a collaborative manner via contrastive learning. Furthermore, a graph mending strategy is utilized to predict missing links in the graph on the server, thus leveraging the capabilities of graph neural networks over high-order graphs. Extensive experiments were conducted on two public datasets, and the results demonstrate the effectiveness of the proposed method.

In Location-based Social Networks (LBSNs), Point-of-Interest (POI) recommendation helps users discover interesting places. There is a trend to move from the conventional cloud-based model to on-device recommendations for privacy protection and reduced server reliance. Due to the scarcity of local user-item interactions on individual devices, solely relying on local instances is not adequate. Collaborative Learning (CL) emerges to promote model sharing among users. Central to this CL paradigm is reference data, which is an intermediary that allows users to exchange their soft decisions without directly sharing their private data or parameters, ensuring privacy and benefiting from collaboration. While recent efforts have developed CL-based POI frameworks for robust and privacy-centric recommendations, they typically use a single and unified reference for all users. Reference data that proves valuable for one user might be harmful to another, given the wide range of user preferences. Some users may not offer meaningful soft decisions on items outside their interest scope. Consequently, using the same reference data for all collaborations can impede knowledge exchange and lead to sub-optimal performance. To address this gap, we introduce the Decentralized Collaborative Learning with Adaptive Reference Data (DARD) framework, which crafts adaptive reference data for effective user collaboration. It first generates a desensitized public reference data pool with transformation and probability data generation methods. For each user, the selection of adaptive reference data is executed in parallel by training loss tracking and influence function. Local models are trained with individual private data and collaboratively with the geographical and semantic neighbors. During the collaboration between two users, they exchange soft decisions based on a combined set of their adaptive reference data. Our evaluations across two real-world datasets highlight DARD's superiority in recommendation performance and addressing the scarcity of available reference data.

In this paper, we present an open-source pure-Python library called PyPop7 for black-box optimization (BBO). As population-based methods (e.g., evolutionary algorithms, swarm intelligence, and pattern search) become increasingly popular for BBO, the design goal of PyPop7 is to provide a unified API and elegant implementations for them, particularly in challenging high-dimensional scenarios. Since these population-based methods easily suffer from the notorious curse of dimensionality owing to random sampling as one of core operations for most of them, recently various improvements and enhancements have been proposed to alleviate this issue more or less mainly via exploiting possible problem structures: such as, decomposition of search distribution or space, low-memory approximation, low-rank metric learning, variance reduction, ensemble of random subspaces, model self-adaptation, and fitness smoothing. These novel sampling strategies could better exploit different problem structures in high-dimensional search space and therefore they often result in faster rates of convergence and/or better qualities of solution for large-scale BBO. Now PyPop7 has covered many of these important advances on a set of well-established BBO algorithm families and also provided an open-access interface to adding the latest or missed black-box optimizers for further functionality extensions. Its well-designed source code (under GPL-3.0 license) and full-fledged online documents (under CC-BY 4.0 license) have been freely available at https://github.com/Evolutionary-Intelligence/pypop and https://pypop.readthedocs.io, respectively.

Gradient inversion attacks (GIAs) have posed significant challenges to the emerging paradigm of distributed learning, which aims to reconstruct the private training data of clients (participating parties in distributed training) through the shared parameters. For counteracting GIAs, a large number of privacy-preserving methods for distributed learning scenario have emerged. However, these methods have significant limitations, either compromising the usability of global model or consuming substantial additional computational resources. Furthermore, despite the extensive efforts dedicated to defense methods, the underlying causes of data leakage in distributed learning still have not been thoroughly investigated. Therefore, this paper tries to reveal the potential reasons behind the successful implementation of existing GIAs, explore variations in the robustness of models against GIAs during the training process, and investigate the impact of different model structures on attack performance. After these explorations and analyses, this paper propose a plug-and-play GIAs defense method, which augments the training data by a designed vicinal distribution. Sufficient empirical experiments demonstrate that this easy-to-implement method can ensure the basic level of privacy without compromising the usability of global model.

Crowd navigation path planning is important in public scenes. Existing strategies are mainly based on manual design, which is not flexible or effective enough. This article proposes an evolutionary framework for automatic crowd navigation path planning in public scenes. The proposed framework contains a new fitness evaluation mechanism that can quantitatively evaluate the quality of a path planning strategy by considering both crowd safety and flow speed. Based on the fitness evaluation mechanism, a framework based on multiobjective differential evolution (DE) is developed to efficiently evolve path planning strategies. Simulation results on two synthetic scenes and a real-world metro station scene show that the proposed framework can provide good path planning strategies.

Privacy-preserving machine learning requires that models do not reveal any private information about their training data. However, model inversion attacks (MIAs), which aim to recover the features of training data, pose a huge threat to the security of AI models. Most existing MIAs assume that the target model is white-box, but most models deployed in reality are black-box, and these models can only be accessed like an oracle. There are a few studies for black-box scenarios, but their performance is limited. In this paper, we firstly formulate the MIA problem completely in Bayesian perspective. Second, we propose a novel two-stage MIA approach, the Coarse-to-Fine Model Inversion Attack (C2FMI), which efficiently addresses the MIA problem in the black-box scenario. In stage I of C2FMI, we design a reverse network that constrains the recovered images (also named attacked images) to fall near the manifold space of the training data. In stage II, we design a black-box oriented strategy which further facilitates the attacked images to approach the training data. Empirically, C2FMI achieves a performance that even surpasses existing white-box attack methods. Furthermore, we design the stability analysis method for analyzing the stability of C2FMI along with existing MIAs. Finally, we explore the potential countermeasures which could defend against our attacks.

Model inversion attacks aim to reveal information about sensitive training data of AI models, which may lead to serious privacy leakage. However, existing attack methods have limitations in reconstructing training data with higher feature fidelity. In this paper, we propose an evolutionary model inversion attack approach (EvoMI) and empirically demonstrate that combined with the systematic search in the multi-degree-of-freedom latent space of the generative model, the simple use of an evolutionary algorithm can effectively improve the attack performance. Concretely, at first, we search for latent vectors which can generate images close to the attack target in the latent space with low-degree of freedom. Generally, the low-freedom constraint will reduce the probability of getting a local optima compared to existing methods that directly search for latent vectors in the high-freedom space. Consequently, we introduce a mutation operation to expand the search domain, thus further reduce the possibility of obtaining a local optima. Finally, we treat the searched latent vectors as the initial values of the post-processing and relax the constraint to further optimize the latent vectors in a higher-freedom space. Our proposed method is conceptually simple and easy to implement, yet it achieves substantial improvements and outperforms the state-of-the-art methods significantly.

To address privacy concerns and reduce network latency, there has been a recent trend of compressing cumbersome recommendation models trained on the cloud and deploying compact recommender models to resource-limited devices for the real-time recommendation. Existing solutions generally overlook device heterogeneity and user heterogeneity. They require devices with the same budget to share the same model and assume the available device resources (e.g., memory) are constant, which is not reflective of reality. Considering device and user heterogeneities as well as dynamic resource constraints, this paper proposes a Personalized Elastic Embedding Learning framework (PEEL) for the on-device recommendation, which generates Personalized Elastic Embeddings (PEEs) for devices with various memory budgets in a once-for-all manner, adapting to new or dynamic budgets, and addressing user preference diversity by assigning personalized embeddings for different groups of users. Specifically, it pretrains a global embedding table with collected user-item interaction instances and clusters users into groups. Then, it refines the embedding tables with local interaction instances within each group. PEEs are generated from the group-wise embedding blocks and their weights that indicate the contribution of each embedding block to the local recommendation performance. Given a memory budget, PEEL efficiently generates PEEs by selecting embedding blocks with the largest weights, making it adaptable to dynamic memory budgets on devices. Furthermore, a diversity-driven regularizer is implemented to encourage the expressiveness of embedding blocks, and a controller is utilized to optimize the weights. Extensive experiments are conducted on two public datasets, and the results show that PEEL yields superior performance on devices with heterogeneous and dynamic memory budgets.

In the post-Moore era, main performance gains of black-box optimizers are increasingly depending on parallelism, especially for large-scale optimization (LSO). Here we propose to parallelize the well-established covariance matrix adaptation evolution strategy (CMA-ES) and in particular its one latest LSO variant called limited-memory CMA-ES (LM-CMA). To achieve efficiency while approximating its powerful invariance property, we present a multilevel learning-based meta-framework for distributed LM-CMA. Owing to its hierarchically organized structure, Meta-ES is well-suited to implement our distributed meta-framework, wherein the outer-ES controls strategy parameters while all parallel inner-ESs run the serial LM-CMA with different settings. For the distribution mean update of the outer-ES, both the elitist and multi-recombination strategy are used in parallel to avoid stagnation and regression, respectively. To exploit spatiotemporal information, the global step-size adaptation combines Meta-ES with the parallel cumulative step-size adaptation. After each isolation time, our meta-framework employs both the structure and parameter learning strategy to combine aligned evolution paths for CMA reconstruction. Experiments on a set of large-scale benchmarking functions with memory-intensive evaluations, arguably reflecting many data-driven optimization problems, validate the benefits (e.g., effectiveness w.r.t. solution quality, and adaptability w.r.t. second-order learning) and costs of our meta-framework.

How to train a generalizable meta-policy by continually learning a sequence of tasks? It is a natural human skill yet challenging to achieve by current reinforcement learning: the agent is expected to quickly adapt to new tasks (plasticity) meanwhile retaining the common knowledge from previous tasks (stability). We address it by "Continual Task Allocation via Sparse Prompting (CoTASP)", which learns over-complete dictionaries to produce sparse masks as prompts extracting a sub-network for each task from a meta-policy network. CoTASP trains a policy for each task by optimizing the prompts and the sub-network weights alternatively. The dictionary is then updated to align the optimized prompts with tasks’ embedding, thereby capturing tasks’ semantic correlations. Hence, relevant tasks share more neurons in the meta-policy network due to similar prompts while cross-task interference causing forgetting is effectively restrained. Given a meta-policy and dictionaries trained on previous tasks, new task adaptation reduces to highly efficient sparse prompting and sub-network finetuning. In experiments, CoTASP achieves a promising plasticity-stability trade-off without storing or replaying any past tasks’ experiences. It outperforms existing continual and multi-task RL methods on all seen tasks, forgetting reduction, and generalization to unseen tasks.

Collaborative filtering (CF) based recommender systems are typically trained based on personal interaction data (e.g., clicks and purchases) that could be naturally represented as ego graphs. However, most existing recommendation methods collect these ego graphs from all users to compose a global graph to obtain high-order collaborative information between users and items, and these centralized CF recommendation methods inevitably lead to a high risk of user privacy leakage. Although recently proposed federated recommendation systems can mitigate the privacy problem, they either restrict the on-device local training to an isolated ego graph or rely on an additional third-party server to access other ego graphs resulting in a cumbersome pipeline, which is hard to work in practice. In addition, existing federated recommendation systems require resource-limited devices to maintain the entire embedding tables resulting in high communication costs.

In many clustering scenes, data samples' attribute values change over time. For such data, we are often interested in obtaining a partition for each time step and tracking the dynamic change of partitions. Normally, a smooth change is assumed for data to have a temporal smooth nature. Existing algorithms consider the temporal smoothness as an a priori preference and bias the search toward the preferred direction. This a priori manner leads to a risk of converging to an unexpected region because it is not always the case that a reasonable preference can be elicited given the little prior knowledge about the data. To address this issue, this article proposes a new clustering framework called evolutionary robust clustering over time. One significant innovation of the proposed framework is processing the temporal smoothness in an a posteriori manner, which avoids unexpected convergence that occurs in existing algorithms. Furthermore, the proposed framework automatically infers the a posteriori preference to temporal smoothness without data's affinity matrix and predefined parameters, which holds better applicability and efficiency. The effectiveness and efficiency of the proposed framework are confirmed by comparing with state-of-the-art algorithms on both synthetic and real datasets.

Community detection in multiplex networks is an emerging research topic in the field of network science. Existing methods usually ignore the similarities among component layers of a multiplex network when detecting its community structures, which decreases the detection efficiency. In this article, we decompose the community detection in multiplex networks into two problems and propose a novel algorithm that can detect both the specific community partition for each component layer (layer-level community structure) and the composite community structure shared by all layers. First, by specifying the modularity optimization on a network layer as an optimization task, we model the layer-level community detection as a multitask optimization (MTO) problem and employ an evolutionary MTO algorithm to solve it. In this way, the topology correlations among different layers can be utilized to facilitate the community detection. Second, we propose an evolutionary clustering ensemble method to find the composite community structure based on the layer-level community partitions and the multiplex network. The proposed method is tested on both synthetic and real-world benchmark networks and compared with classical and state-of-the-art algorithms. Experimental results show that the proposed algorithm has superior community detection performances on multiplex networks.

Recommender systems play a significant role in information filtering and have been utilized in different scenarios, such as e-commerce and social media. With the prosperity of deep learning, deep recommender systems show superior performance by capturing non-linear information and item-user relationships. However, the design of deep recommender systems heavily relies on human experiences and expert knowledge. To tackle this problem, Automated Machine Learning (AutoML) is introduced to automatically search for the proper candidates for different parts of deep recommender systems. This survey performs a comprehensive review of the literature in this field. Firstly, we propose an abstract concept for AutoML for deep recommender systems (AutoRecSys) that describes its building blocks and distinguishes it from conventional AutoML techniques and recommender systems. Secondly, we present a taxonomy as a classification framework containing feature selection search, embedding dimension search, feature interaction search, model architecture search, and other components search. Furthermore, we put a particular emphasis on the search space and search strategy, as they are the common thread to connect all methods within each category and enable practitioners to analyze and compare various approaches. Finally, we propose four future promising research directions that will lead this line of research.

As a special class of multiobjective optimization problems (MOPs), multiparty multiobjective optimization prob-lems (MPMOPs) widely exist in real-world applications. In MPMOPs, there are multiple decision makers (DMs) concerning multiple different conflicting objectives. The goal of solving MPMOPs is to catch the best solutions satisfying all DMs as far as possible. To our best knowledge, there is little attention on solving MPMOPs, and only two optimization algorithms, i.e., OptMPNDS and OptMPNDS2, are proposed. These two algorithms are both based on non-dominated sorting genetic algorithm II (NSGA-II). However, there is no algorithm pro-posed from the decomposition perspective to solve MPMOPs. Multiobjective evolutionary algorithm based on decomposition (MOEA/D) is a popular multiobjective evolutionary optimization algorithm for MOPs. In this paper, we embed the party-by-party strategy into MOEA/D and propose the novel optimization algorithm MOEA/D-MP to solve MPMOPs. The experimental results on the benchmarks have demonstrated the effectiveness of MOEA/D-MP.

As a crucial component of most modern deep recommender systems, feature embedding maps high-dimensional sparse user/item features into low-dimensional dense embeddings. However, these embeddings are usually assigned a unified dimension, which suffers from the following issues: (1) high memory usage and computation cost. (2) sub-optimal performance due to inferior dimension assignments. In order to alleviate the above issues, some works focus on automated embedding dimension search by formulating it as hyper-parameter optimization or embedding pruning problems. However, they either require well-designed search space for hyperparameters or need time-consuming optimization procedures. In this paper, we propose a Single-Shot Embedding Dimension Search method, called SSEDS, which can efficiently assign dimensions for each feature field via a single-shot embedding pruning operation while maintaining the recommendation accuracy of the model. Specifically, it introduces a criterion for identifying the importance of each embedding dimension for each feature field. As a result, SSEDS could automatically obtain mixed-dimensional embeddings by explicitly reducing redundant embedding dimensions based on the corresponding dimension importance ranking and the predefined parameter budget. Furthermore, the proposed SSEDS is model-agnostic, meaning that it could be integrated into different base recommendation models. The extensive offline experiments are conducted on two widely used public datasets for CTR (Click Through Rate) prediction task, and the results demonstrate that SSEDS can still achieve strong recommendation performance even if it has reduced 90% parameters. Moreover, SSEDS has also been deployed on the WeChat Subscription platform for practical recommendation services. The 7-day online A/B test results show that SSEDS can significantly improve the performance of the online recommendation model while reducing resource consumption.

Unmanned Aerial Vehicles (UAVs), or drones, have the potential to be applied to delivery services, which are expected to bring economic benefits. One of the key issues is planning routes for vehicles and drones with specific constraints and objectives, known as Vehicle Routing Problem with Drones (VRPD). This paper considers a scenario involving multiple trucks, multiple UAV stations, and UAVs within each station to serve the customers. A bi-level approach that combines the Brain Storm optimization algorithm and Adaptive Large Neighborhood Search is proposed by designing the solution representation, new solution generation mechanism, and other operations. The experimental results show that the proposed method has the ability to solve the problem and deserves further development.

Dynamic multiobjective optimization problems are challenging due to their fast convergence and diversity maintenance requirements. Prediction-based evolutionary algorithms currently gain much attention for meeting these requirements. However, it is not always the case that an elaborate predictor is suitable for different problems and the quality of historical solutions is sufficient to support prediction, which limits the availability of prediction-based methods over various problems. Faced with these issues, this article proposes a knowledge learning strategy for change response in the dynamic multiobjective optimization. Unlike prediction approaches that estimate the future optima from previously obtained solutions, in the proposed strategy, we react to changes via learning from the historical search process. We introduce a method to extract the knowledge within the previous search experience. The extracted knowledge can accelerate convergence as well as introduce diversity for the optimization of the future environment. We conduct a comprehensive experiment on comparing the proposed strategy with the state-of-the-art algorithms. Results demonstrate the better performance of the proposed strategy in terms of solution quality and computational efficiency.

The multiple-preys pursuit (MPP) is the adversarial game between predators and preys. If the capture of a prey is defined as that it cannot move anymore due to the surrounding of predators, there are two kinds of task allocations. One is about assigning which prey to which group of predators so that all preys can be captured. The other is about assigning which capturing position to which predator to encircle the prey simultaneously. In this paper, the MPP is modeled as a dynamic optimization problem and each its time step is solved in two stages. Firstly, the first kind of task allocation problem is modeled as the biquadratic assignment problem (BiQAP) and a MPP fitness function is proposed for the evaluation of such BiQAP task allocations. In this way, the MPP is transformed to several single-prey pursuit (SPP) problems. Secondly, for each SPP, we extend the coordinated SPP strategy CCPSO-R (cooperative coevolutionary particle swarm optimization for robots) to its parallel version as PCCPSO-R to enable the parallel implicit capturing position allocating by parallel observation, decision making, and moving of predators. Through experiments of the current BiQAP solvers on the task allocation, we improve the best one of them in statistic based on the domain knowledge. Moreover, the advantages of PCCPSO-R in the capturing efficiency over CCPSO-R is testified in the MPP experiments.

Population-based methods are often used to solve multimodal optimization problems. By combining niching or clustering strategy, the state-of-the-art approaches generally divide the population into several subpopulations to find multiple solutions for a problem at hand. However, these methods only guided by the fitness value during iterations, which are suffering from determining the number of subpopulations, i.e., the number of niche areas or clusters. To compensate for this drawback, this paper presents an Attention-oriented Brain Storm Optimization (ABSO) method that introduces the attention mechanism into a relatively new swarm intelligence algorithm, i.e., Brain Storm Optimization (BSO). By converting the objective space from the fitness space into "attention" space, the individuals are clustered and updated iteratively according to their salient values. Rather than converge to a single global optimum, the proposed method can guide the search procedure to converge to multiple "salient" solutions. The preliminary results show that the proposed method can locate multiple global and local optimal solutions of several multimodal benchmark functions. The proposed method needs less prior knowledge of the problem and can automatically converge to multiple optimums guided by the attention mechanism, which has excellent potential for further development.

Imbalanced classification on graphs is ubiquitous yet challenging in many real-world applications, such as fraudulent node detection. Recently, graph neural networks (GNNs) have shown promising performance on many network analysis tasks. However, most existing GNNs have almost exclusively focused on the balanced networks, and would get unappealing performance on the imbalanced networks. To bridge this gap, in this paper, we present a generative adversarial graph network model, called ImGAGN to address the imbalanced classification problem on graphs. It introduces a novel generator for graph structure data, named GraphGenerator, which can simulate both the minority class nodes' attribute distribution and network topological structure distribution by generating a set of synthetic minority nodes such that the number of nodes in different classes can be balanced. Then a graph convolutional network (GCN) discriminator is trained to discriminate between real nodes and fake (i.e., generated) nodes, and also between minority nodes and majority nodes on the synthetic balanced network. To validate the effectiveness of the proposed method, extensive experiments are conducted on four real-world imbalanced network datasets. Experimental results demonstrate that the proposed method ImGAGN outperforms state-of-the-art algorithms for semi-supervised imbalanced node classification task.

The local community detection is a significant branch of the community detection problems. It aims at finding the local community to which a given starting node belongs. The local community detection plays an important role in analyzing the complex networks and recently has drawn much attention from the researchers. In the past few years, several local community detection algorithms have been proposed. However, the previous methods only make use of the limited local information of networks but overlook the other valuable information. In this article, we propose an evolutionary computation-based algorithm called evolutionary-based local community detection (ELCD) algorithm to detect local communities in the complex networks by taking advantages of the entire obtained information. The performance of the proposed algorithm is evaluated on both synthetic and real-world benchmark networks. The experimental results show that the proposed algorithm has a superior performance compared with the state-of-the-art local community detection methods. Furthermore, we test the proposed algorithm on incomplete real-world networks to show its effectiveness on the networks whose global information cannot be obtained.

In brain storm optimization (BSO), the convergent operation utilizes a clustering strategy to group the population into multiple clusters, and the divergent operation uses this cluster information to generate new individuals. However, this mechanism is inefficient to regulate the exploration and exploitation search. This article first analyzes the main factors that influence the performance of BSO and then proposes an orthogonal learning framework to improve its learning mechanism. In this framework, two orthogonal design (OD) engines (i.e., exploration OD engine and exploitation OD engine) are introduced to discover and utilize useful search experiences for performance improvements. In addition, a pool of auxiliary transmission vectors with different features is maintained and their biases are also balanced by the OD decision mechanism. Finally, the proposed algorithm is verified on a set of benchmarks and is adopted to resolve the quantitative association rule mining problem considering the support, confidence, comprehensibility, and netconf. The experimental results show that the proposed approach is very powerful in optimizing complex functions. It not only outperforms previous versions of the BSO algorithm but also outperforms several famous OD-based algorithms.

The Vehicle Routing Problem with Time Windows (VRPTW) is NP-hard which has many real-world applications in logistics and transportation. The traditional VRPTW is defined on a complete graph with customers as nodes, but in the real-world, VRPTWs are more based on road networks. To better simulate the real-world scenarios, this paper studies the VRPTW on road networks. Most researchers solve the VRPTW on road networks by utilizing exact algorithms which can not deal with large size problem. In this paper, a hybrid BSO-ACS algorithm, which combines Brain Storm optimization (BSO), Ant Colony System (ACS) and Local Search (LS), is proposed to solve the VRPTW on road networks. A set of instances based on the road network of southwest Shenzhen, China are generated as benchmark problems. The computational experiments demonstrate the effectiveness of the proposed algorithm.

Swarm intelligence algorithms have been widely and successfully used to optimize many science and engineering problems, the collective behavior of the agents lead to the emergence of intelligence. These interactions among agents can be classified into three categories: exploring, emulating and learning. Brain Storm Optimization (BSO) is a novel swarm intelligence algorithm which is inspired by the human brainstorming process, and generates new ideas by emulating existing ideas. In this paper, a new BSO algorithm with an adaptive learning strategy (BSOAL) is proposed. By considering the evolutionary speed factor of each individual and the aggregation degree of the swarm, the proposed BSO-AL generates new individuals by exploring, emulating or learning adaptively. Comparative experiments were conducted on a set of benchmark functions with different dimensions. The experimental results show that the proposed BSO-AL algorithm outperforms the classic BSO algorithm and the other two state-of-the-art algorithms, which demonstrates the effectiveness of the learning strategy.

Designing the collaborative mechanism is a fundamental problem for the multi-robot systems. It aims to determine the perception, communication, and motion strategies for a single robot to obtain the desired behavior at the system level. Generally, we can use manual design and automatic design, or the combination of the above two approaches for particular system behavior. With the increasing of task complexity and the uncertainty of surroundings, the adaptability and autonomy are hard to achieve with manual design approaches. By using the mechanism of learning or evolution, automatic design can generate sensor configurations, communication parameters, as well as control strategies automatically, which has been widely concerned in recent years. In this paper, the brainstorming method of collaborative problem-solving in human society is introduced into the design of multi-robot systems. This paper proposes an automatic design framework: Brain Storm Robotics(BSR), in which the system architecture, the representation of ideas, and the generation of new ideas are discussed. The effectiveness of the proposed BSR framework is verified by an example of designing an aggregation behavior for a swarm of robots. The results show that the control strategy designed by this framework is more efficient than that designed manually, which has outstanding development prospects. The future researches for the development of this potential framework are also discussed.

Recently, graph neural networks (GNNs) have been shown to be an effective tool for learning the node representations of the networks and have achieved good performance on the semi-supervised node classification task. However, most existing GNNs methods fail to take networks’ temporal information into account, therefore, cannot be well applied to dynamic network applications such as the continuous-time link prediction task. To address this problem, we propose a Temporal Dependent Graph Neural Network (TDGNN), a simple yet effective dynamic network representation learning framework which incorporates the network temporal information into GNNs. TDGNN introduces a novel Temporal Aggregator (TDAgg) to aggregate the neighbor nodes’ features and edges’ temporal information to obtain the target node representations. Specifically, it assigns the neighbor nodes aggregation weights using an exponential distribution to bias different edges’ temporal information. The performance of the proposed method has been validated on six real-world dynamic network datasets for the continuous-time link prediction task. The experimental results show that the proposed method outperforms several state-of-the-art baselines.

This article presents a limit-cycle-based mutant multiobjective pigeon-inspired optimization (PIO). In this algorithm, the limit-cycle-based mechanism is devised to consider the factors that affect the flight of pigeons to simplify the multiobjective PIO algorithm. The mutant mechanism is incorporated to strengthen the exploration capability in the evolutionary process. Additionally, the application of the dual repository makes the nondominated solutions stored and selected to guide the flight of pigeons. Attributed to the limit-cycle-based mutant mechanisms, this algorithm not only obtains the faster convergence speed and higher accuracy but also improves its population diversity. To confirm the universal application of this algorithm, theoretical analysis of the convergence is discussed in this article. Finally, comparative experiments of our proposed algorithm and other five multiobjective methods are conducted to verify the accuracy, efficiency, and convergence stability of the proposed algorithm.

The benchmark problems have played a fundamental role in verifying the algorithm's search ability. A dynamic multimodal optimization (DMO) problem is defined as an optimization problem with multiple global optima and characteristics of global optima which are changed during the search process. Two cases are used to illustrate the application scenario of DMO. A set of benchmark functions on DMO, which contains eight problems, are proposed to show the difficulty of DMO. The properties of the proposed benchmark problems, such as the distribution of solutions, the scalability, the number of global/local optima, are discussed.

Decomposition-based optimizers have shown very promising computational and convergence performance on many large-scale real-parameter optimization problems. Among them, a class of recently proposed cooperative coevolutionary algorithms (CCEAs) and a type of conventional block coordinate descent algorithms (BCDAs) are arguably the two most representative frameworks applied to the minimization of non-differentiable and differentiable objective function, respectively. This paper explores the connections between CCEAs and BCDAs, which can help gain deeper understandings of CCEAs. First, we propose a unified analytical framework for both CCEAs and BCDAs to capture the common game-theoretic nature by combining their respective theoretical advances. Second, many real-world objective functions are non-additively separable, where all decision variables interact with each other in a direct or indirect fashion. However, most of the state-of-the-art decomposition strategies for CCEAs can only capture the simple additive separability and cannot recognize the non-additive separability, but which has been widely studied in the BCDAs context. The performance of CCEAs on such functions is yet to be fully understood since intuitively CCEAs seem to be not suitable for them. We use the proposed framework to confirm and extend CCEAs' applicability to a special class of non-additively separable functions. Finally, based on the proposed framework, we provide two practical suggestions as well as a suite of new test functions to help design practically better CCEAs for large-scale optimization.

Parameter control is critical to the performance of any evolutionary algorithm (EA). In this paper, we propose a Q-Learning-based Particle Swarm Optimization (QLPSO) algorithm, which uses the Reinforcement Learning (RL) to train the parameters in Particle Swarm Optimization (PSO) algorithm. The core of the QLPSO algorithm is a three-dimensional Q table which consists of a state plane and an action axis. The state plane includes the state of the particles in both of the decision space and the objective space. The action axis controls the exploration and exploitation of particles by setting different parameters. The Q table can help particles to select actions according to their states. Besides, the Q table should be updated by reward function which is designed according to the performance change of particles and the number of iterations. The main difference between the QLPSO algorithms for single-objective and multi-objective optimization lies in the evaluation of the solution performance. In single-objective optimization, we only compare the fitness values of solutions, while in multi-objective optimization, we need to discuss the dominant relationship between solutions with the help of Pareto front. The performance of QLPSO is tested based on 6 single-objective and 5 multi-objective benchmark functions. The experiment results reveal the competitive performance of QLPSO compared with other algorithms.

The brain storm optimization (BSO) algorithm is a new and promising swarm intellgience paradigm, inspired from the behaviors of the human process of brainstorming. The noverty of BSO lies in the clustering mechanism where the ideas are clustered into a set of groups and each idea learns from experiences of one inter-cluster or two intra-cluster neighbors. However, this mechanism is inefficient to deal with complex optimiaiton problems. In this paper, we propose an improved BSO algorithm called OSBSO using orthogonal experimental design (OED) strategy, which aims to discover useful search experiences for improving the convergence and solution accurancy. In OSBSO, two new clustering procedures are developed, i.e., orthogonal initialization and orthogonal clustering. The orthogonal initialization aims to improve the uniformity of the initial cluster centers in the objective space instead of the decision space, which can enhance the convergence performance. The orthogonal clustering uses the information between inter- cluster and intra-cluster indviduals to alleviate the evolution stagnation of clusters. Experiments are conducted on a set of the CEC2017 benchmark functions and the results verify the effectivenss and efficiency of OSBSO.

Autonomous aerial refueling (AAR) is a crucial technique of unmanned aerial vehicles (UAVs) to push the fuel limits and play a great role in both civilian and military domains. This paper presents an accurate and robust binocular pose estimation algorithms optimized by brain storm optimization (BSO), which is developed from a robust non-iterative solution of PnP (RPnP). In this algorithm, BSO is employed to select the best rotation axis in RPnP. A large quantity of contrastive simulation experiments has been conducted to verify the proposed algorithm. Furthermore, this work built an aerial verification platform for vision-based AAR. A tanker UAV and a receiver UAV were applied to implement AAR. The real-time visual measuring system includes feature extraction and pose estimation. Several state-of-the-art pose estimation algorithms and the proposed method (which refers to BSO-BPnP) have been tested in the aerial verification platform. Adequate comparative trials and detailed analyses are given in this paper.

The seven papers in this special section were presented at the ICSI 2016 Conference. These articles are primarily dealing with either novel bioinspired swarm intelligence algorithms and their improvements aswell as some practical applications inmulti-objective optimization, network community detection, curve fitting, and swarm robotics, etc.

The development, implementation, variant, and future directions of a new swarm intelligence algorithm, brain storm optimization (BSO) algorithm, are comprehensively surveyed. Brain storm optimization algorithm is a new and promising swarm intelligence algorithm, which simulates the human brainstorming process. Through the convergent operation and divergent operation, individuals in BSO are grouped and diverged in the search space/objective space. To the best of our knowledge, there are 75 papers, 8 theses, and 5 patents in total on the development and application of the BSO algorithm. Every individual in the BSO algorithm is not only a solution to the problem to be optimized, but also a data point to reveal the landscape of the problem. Based on the developments of brain storm optimization algorithms, different kinds of optimization problems and real-world applications could be solved.

Recently, inspired by the human brainstorming process, a new kind of metaheuristic algorithm, called brain storm optimization (BSO) algorithm was proposed for global optimization. Experimental results have shown its excellent performance when solving optimization problems. In order to further improve the search ability of the BSO, this paper proposes an improved BSO (IBSO) algorithm by introducing graph theory into it. In IBSO, new individuals will be generated to replace some old individuals when the BSO algorithm is in a poor status. Whether a BSO algorithm is in a poor status is determined by the length of Hamiltonian cycle, which can be obtained by transferring all the individuals into an undirected weight graph. A Hamiltonian cycle and its length will be computed according to a modified cycle algorithm. The proposed IBSO algorithm is tested on twelve benchmarks, and the experimental results illustrate its effectiveness.

Presents the introductory editorial for this issue of the publication.

Brain storm optimization (BSO) algorithm is a new kind of swarm intelligence algorithm, which is inspired by collective behavior of human beings. In this paper, a Markov model for brain storm optimization algorithm is derived. The model gives the theoretical probability of the occurrence of each possible population as the number of generation count goes to infinity. Using the Markov model, the convergence of the brain storm optimization is analyzed.

The learning strategy in the canonical particle swarm optimization (PSO) algorithm is often blamed for being the primary reason for loss of diversity. Population diversity maintenance is crucial for preventing particles from being stuck into local optima. In this paper, we present an improved PSO algorithm with an interswarm interactive learning strategy (IILPSO) by overcoming the drawbacks of the canonical PSO algorithm's learning strategy. IILPSO is inspired by the phenomenon in human society that the interactive learning behavior takes place among different groups. Particles in IILPSO are divided into two swarms. The interswarm interactive learning (IIL) behavior is triggered when the best particle's fitness value of both the swarms does not improve for a certain number of iterations. According to the best particle's fitness value of each swarm, the softmax method and roulette method are used to determine the roles of the two swarms as the learning swarm and the learned swarm. In addition, the velocity mutation operator and global best vibration strategy are used to improve the algorithm's global search capability. The IIL strategy is applied to PSO with global star and local ring structures, which are termed as IILPSO-G and IILPSO-L algorithm, respectively. Numerical experiments are conducted to compare the proposed algorithms with eight popular PSO variants. From the experimental results, IILPSO demonstrates the good performance in terms of solution accuracy, convergence speed, and reliability. Finally, the variations of the population diversity in the entire search process provide an explanation why IILPSO performs effectively.

Cooperative Coevolution (CC) was introduced into evolutionary algorithms as a promising framework for tackling large scale optimization problems through a divide-and-conquer strategy. A number of decomposition methods to identify interacting variables have been proposed to construct subcomponents of a large scale problem, but if the variables are all non-separable, all the CC-based algorithms of decomposition will lose the functionality, therefore, classical CC-based algorithms are inefficient in processing non-separable problems that have many interacting variables. In this paper, a new CC framework which integrates global and local search algorithms is proposed for solving large scale optimization problems. In the stage of global cooperative coevolution, we introduce a new interacting variables grouping method named Sequential Sliding Window. When the performance of global search reaches a deviation tolerance or the variables are fully non-separable, we then use a more effective local search algorithm to subsequently search the solution space of the large scale optimization problem. The integration of global and local algorithms into CC framework can efficiently improve the capability in processing large scale non-separable problems. Experimental results on large scale optimization benchmarks show that the proposed framework is more effective than other existing CC frameworks.

The aim of multimodal optimization is to locate multiple peaks/optima in a single run and to maintain these found optima until the end of a run. The results of seven variants of particle swarm optimization (PSO) algorithms on IEEE Congress on Evolutionary Computation (CEC) 2015 single objective multi-niche optimization problems are reported in this paper. The PSO algorithms include PSO with star structure, PSO with ring structure, PSO with four clusters structure, PSO with Von Neumann structure, social-only PSO with star structure, social-only PSO with ring structure, and cognition-only PSO. The experimental tests are conducted on fifteen benchmark functions. Based on the experimental results, the conclusions could be made that the PSO with ring structure performs better than the other PSO variants on multimodal optimization. To obtain good performance on the multimodal optimization problems, an algorithm needs to converge the candidate solutions to the global optima while keep the population diversity during whole search process.

Brain storm optimization algorithm is a newly proposed swarm intelligence algorithm, which has two main operations, i.e., convergent operation and divergent operation. In the original brain storm optimization algorithm, a clustering algorithm is utilized to cluster individuals into clusters as the convergent operation which is time consuming because of distance calculation during clustering. In this paper, a new convergent operation is proposed to be implemented in the 1-dimensional objective space instead of in the solution space. As a consequence, its computation time will depend on only the population size, not the problem dimension, therefore, a big computation time saving can be obtained which makes it have good scalability. Experimental results demonstrate the effectiveness and efficiency of the proposed brain storm optimization algorithm in objective space.

Swarm intelligence suffers the premature convergence, which happens partially due to the solutions getting clustered together, and not diverging again. The brain storm optimization (BSO), which is a young and promising algorithm in swarm intelligence, is based on the collective behavior of human being, that is, the brainstorming process. Premature convergence also happens in the BSO algorithm. The solutions get clustered after a few iterations, which indicate that the population diversity decreases quickly during the search. A definition of population diversity in BSO algorithm to measure the change of solutions' distribution is proposed in this paper. The algorithm's exploration and exploitation ability can be measured based on the change of population diversity. Two kinds of partial re-initialization strategies are utilized to improve the population diversity in BSO algorithm. The experimental results show that the performance of the BSO is improved by these two strategies.

In swarm intelligence algorithms, premature convergence happens partially due to the solutions getting clustered together, and not diverging again. However, solution clustering is not always harmful for optimization. The solution clustering strategy is utilized in brain storm optimization (BSO) to guide individuals to move toward the better and better areas. The information of clusters indicates the solutions' distribution in the search space, which could be utilized to reveal the landscapes and other proprieties of problems being optimized. In this paper, the solution clustering, and other properties of the brain storm optimization algorithm are analyzed and discussed. Experimental results show that brain storm optimization is a very promising algorithm for solving different kinds of problems.

In this paper, the nearest neighbor method on Chinese text categorization is formulated as an optimization problem. The particle swarm optimization is utilized to optimize a nearest neighbor classifier to solve the Chinese text categorization problem. The parameter k was first optimized to obtain the minimum error, then the categorization problem is formulated as a discrete, constrained, and single objective optimization problem. Each dimension of solution vector is dependent on each other in the solution space. The parameter k and the number of labeled examples for each class are optimized together to reach the minimum categorization error. In the experiment, with the utilization of particle swarm optimization, the performance of a nearest neighbor algorithm can be improved, and the algorithm can obtain the minimum categorization error rate.

Traditional multiobjective evolutionary algorithms (MOEAs) consider multiple objectives as a whole when solving multiobjective optimization problems (MOPs). However, this consideration may cause difficulty to assign fitness to individuals because different objectives often conflict with each other. In order to avoid this difficulty, this paper proposes a novel coevolutionary technique named multiple populations for multiple objectives (MPMO) when developing MOEAs. The novelty of MPMO is that it provides a simple and straightforward way to solve MOPs by letting each population correspond with only one objective. This way, the fitness assignment problem can be addressed because the individuals' fitness in each population can be assigned by the corresponding objective. MPMO is a general technique that each population can use existing optimization algorithms. In this paper, particle swarm optimization (PSO) is adopted for each population, and coevolutionary multiswarm PSO (CMPSO) is developed based on the MPMO technique. Furthermore, CMPSO is novel and effective by using an external shared archive for different populations to exchange search information and by using two novel designs to enhance the performance. One design is to modify the velocity update equation to use the search information found by different populations to approximate the whole Pareto front (PF) fast. The other design is to use an elitist learning strategy for the archive update to bring in diversity to avoid local PFs. CMPSO is comprehensively tested on different sets of benchmark problems with different characteristics and is compared with some state-of-the-art algorithms. The results show that CMPSO has superior performance in solving these different sets of MOPs.

Particle swarm optimization (PSO) may lose search efficiency when the problem's dimension increases to large scale. For high dimensional search space, an algorithm may not be easy to locate at regions which contain good solutions. The exploitation ability is also reduced due to high dimensional search space. The “No Free Lunch” theorem implies that we can make better algorithm if an algorithm knows the information of the problem. Algorithms should have an ability of learning to solve different problems, in other words, algorithms can adaptively change to suit the landscape of problems. In this paper, the strategy of dynamical exploitation space reduction is utilized to learn problems' landscapes. While at the same time, partial re-initialization strategy is utilized to enhance the algorithm's exploration ability. Experimental results show that a PSO with these two strategies has better performance than the standard PSO in large scale problems. Population diversities of variant PSOs, which include position diversity, velocity diversity and cognitive diversity, are discussed and analyzed. From diversity analysis, we can conclude that an algorithm's exploitation ability can be enhanced by exploitation space reduction strategy.

For many large scale learning problems, acquiring a large amount of labeled training data is expensive and time-consuming. Semi-supervised learning is a machine learning paradigm which deals with utilizing unlabeled data to build better classifiers. However, unlabeled data with wrong predictions will mislead the classifier. In this paper, we proposed a particle swarm optimization based semi-learning classifier to solve Chinese text categorization problem. This classifier utilizes an iterative strategy, and the result of classifier is determined by a document's previous prediction and its neighbors' information. The new classifier is tested on a Chinese text corpus. The proposed classifier is compared with the k nearest neighbor method, the k weighted nearest neighbor method, and the self-learning classifier.

Premature convergence happens in Particle Swarm Optimization (PSO) partially due to improper search information propagation. Fast propagation of search information will lead particles get clustered together quickly. Determining a proper search information propagation mechanism is important in optimization algorithms to balance between exploration and exploitation. In this paper, we attempt to figure out the relationship between search information propagation and the population diversity change. Firstly, we analyze the different characteristics of search information propagation in PSO with four kinds of topologies: star, ring, four clusters, and Von Neumann. Secondly, population diversities of PSO, which include position diversity, velocity diversity, and cognitive diversity, are utilized to monitor particles' search during optimization process. Position diversity, velocity diversity, and cognitive diversity, represent distributions of current solutions, particles' “moving potential”, and particles' “moving target”, respectively. From the observation of population diversities, the effect of search information propagation on PSO's optimization performance is discussed at last.

With the rapid development of high-throughput DNA sequencing technologies, the amount of DNA sequence data is accumulating exponentially. The huge influx of data creates new challenges for storage and transmission. This paper proposes a novel adaptive particle swarm optimization-based memetic algorithm (POMA) for DNA sequence compression. POMA is a synergy of comprehensive learning particle swarm optimization (CLPSO) and an adaptive intelligent single particle optimizer (AdpISPO)-based local search. It takes advantage of both CLPSO and AdpISPO to optimize the design of approximate repeat vector (ARV) codebook for DNA sequence compression. ARV is first introduced in this paper to represent the repeated fragments across multiple sequences in direct, mirror, pairing, and inverted patterns. In POMA, candidate ARV codebooks are encoded as particles and the optimal solution, which covers the most approximate repeated fragments with the fewest base variations, is identified through the exploration and exploitation of POMA. In each iteration of POMA, the leader particles in the swarm are selected based on weighted fitness values and each leader particle is fine-tuned with an AdpISPO-based local search, so that the convergence of the search in local region is accelerated. A detailed comparison study between POMA and the counterpart algorithms is performed on 29 (23 basic and 6 composite) benchmark functions and 11 real DNA sequences. POMA is observed to obtain better or competitive performance with a limited number of function evaluations. POMA also attains lower bits-per-base than other state-of-the-art DNA-specific algorithms on DNA sequence data. The experimental results suggest that the cooperation of CLPSO and AdpISPO in the framework of memetic algorithm is capable of searching the ARV codebook space efficiently.

Some of the existing relay schemes for cooperative networks based on one-way relaying protocol are not applicable to the case of two-way relaying protocol. In this letter, a cooperative relay scheme for distributed amplify-and-forward relays working under the two-way relaying protocol is designed in a closed-form, which is asymptotically optimal in the high levels of signal-to-noise ratio (SNR). An upper-bound of the mean squared error (MSE) is derived via a tight approximation of the SNR expression. Based on the minimization of this upper-bound, the two-way relay scheme is then derived as a Rayleigh-Ritz ratio problem. The new relay scheme achieves the minimum MSE for both directional transmissions. Simulation results illustrate the effectiveness of the proposed scheme especially in the high SNR regime.

In the field of evolutionary computation, an important attribute of a population is diversity. This paper proposes a method for measuring the diversity of a particle swarm optimization population. It involves the measurement of position and velocity attributes of the particles that comprise the population. The proposed method is computationally straightforward and is adaptable to other evolutionary algorithms.

In orthogonal frequency division multiplexing (OFDM) systems, some subcarriers at the borders of the allocated bandwidth are usually used as guard band. Since these subcarriers, which are often referred to as virtual subcarriers, are not used for transmission, approach of conventional uniformly placed pilot tones is not applicable any more in some situations. Therefore, it is necessary to derive the optimal pilot sequences based on the nonuniform placement of pilot tones. In this paper, we first compute the mean square error (MSE) of the least squares (LS) channel estimate for multiple-input multiple-output (MIMO) OFDM systems. Then, based on nonuniform pilot tone placement, we derive the optimal pilot sequences with respect to this MSE. Simulation results demonstrate the effectiveness of the proposed approach

This paper reviews the development of the particle swarm optimization method in recent years. Included are brief discussions of various parameters. Modifications to adapt to different and complex environments are reviewed, and real world applications are listed.

The paper presents a modified particle swarm optimization (PSO) algorithm for engineering optimization problems with constraints. PSO is started with a group of feasible solutions and a feasibility function is used to check if the newly explored solutions satisfy all the constraints. All the particles keep only those feasible solutions in their memory. Several engineering design optimization problems were tested and the results show that PSO is an efficient and general approach to solve most nonlinear optimization problems with inequity constraints.

This paper presents a modified dynamic neighborhood particle swarm optimization (DNPSO) algorithm for multiobjective optimization problems. PSO is modified by using a dynamic neighborhood strategy, new particle memory updating, and one-dimension optimization to deal with multiple objectives. An extended memory is introduced to store global Pareto optimal solutions to reduce computation time. Several benchmark cases were tested and the results show that the modified DNPSO is much more efficient than the original DNPSO and other multiobjective optimization techniques.

This paper introduces a modified particle swarm optimizer which deals with permutation problems. Particles are defined as permutations of a group of unique values. Velocity updates are redefined based on the similarity of two particles. Particles change their permutations with a random rate defined by their velocities. A mutation factor is introduced to prevent the current pBest from becoming stuck at local minima. Preliminary study on the n-queens problem shows that the modified PSO is promising in solving constraint satisfaction problems.

The use of structured lighting is common in machine vision systems where three-dimensional measurements must be made. In one method, a matrix of bright spots is projected onto an object. The spots move from predetermined positions depending on the distance between the object and the camera. Each spot must be labeled and tracked in order to compute the distance. In this paper, particle swarm optimization (PSO) has been utilized to label a structured beam matrix. Experimental results illustrate the effectiveness of the PSO approach.

This paper focuses on the engineering and computer science aspects of developments, applications, and resources related to particle swarm optimization. Developments in the particle swarm algorithm since its origin in 1995 are reviewed. Included are brief discussions of constriction factors, inertia weights, and tracking dynamic systems. Applications, both those already developed, and promising future application areas, are reviewed. Finally, resources related to particle swarm optimization are listed, including books, Web sites, and software. A particle swarm optimization bibliography is at the end of the paper.

Using particle swarms to track and optimize dynamic systems is described. Issues related to tracking and optimizing dynamic systems are briefly reviewed. Three kinds of dynamic systems are defined for the purposes of this paper. One of them is chosen for preliminary analysis using the particle swarm on the parabolic benchmark function. Successful tracking of a 10-dimensional parabolic function with a severity of up to 1.0 is demonstrated. A number of issues related to tracking and optimizing dynamic systems with particle swarms are identified. Directions for future research and applications are suggested.

A fuzzy system is implemented to dynamically adapt the inertia weight of the particle swarm optimization algorithm (PSO). Three benchmark functions with asymmetric initial range settings are selected as the test functions. The same fuzzy system has been applied to all three test functions with different dimensions. The experimental results illustrate that the fuzzy adaptive PSO is a promising optimization method, which is especially useful for optimization problems with a dynamic environment.

The performance of particle swarm optimization using an inertia weight is compared with performance using a constriction factor. Five benchmark functions are used for the comparison. It is concluded that the best approach is to use the constriction factor while limiting the maximum velocity Vmax to the dynamic range of the variable Xmax on each dimension. This approach provides performance on the benchmark functions superior to any other published results known by the authors.

We empirically study the performance of the particle swarm optimizer (PSO). Four different benchmark functions with asymmetric initial range settings are selected as testing functions. The experimental results illustrate the advantages and disadvantages of the PSO. Under all the testing cases, the PSO always converges very quickly towards the optimal positions but may slow its convergence speed when it is near a minimum. Nevertheless, the experimental results show that the PSO is a promising optimization method and a new approach is suggested to improve PSO's performance near the optima, such as using an adaptive inertia weight.

Chaotic behavior can be found in nonautonomous cellular neural networks (CNNs). The impulsive control method to control this kind of chaos is used and some satisfactory result are achieved. The condition which ensure existence of periodic solution in the impulse controlled system is provided and proved, and numerical simulation shows the chaos can be eliminated by adding a very small external force. Through observing the time waveform diagram of some controlled periodic orbits, we find the time waveform is very similar to encephalic electric activity or cardiac electric activity in biomedical field study. This illustrate that the nonautonomous CNNs model can successfully model physiological electric response activity signals. Moreover, the physiological explanation is given for the nonautonomous CNNs model.

Evolutionary fuzzy systems are discussed in which the membership function shapes and types and the fuzzy rule set including the number of rules inside it are evolved using a genetic (evolutionary) algorithm. In addition, the genetic parameters (operators) of the evolutionary algorithm are adapted via a fuzzy system. Benefits of the methodology are illustrated in the process of classifying the iris data set. Possible extensions of the methods are summarized.