Search

This review aims to summarize and describe research on the topic of automatic group emotion recognition. In recent years, the topic of emotion analysis of groups or crowds has gained interest, with studies performing emotion detection in different contexts, using different datasets and modalities (such as images, video, audio, social media messages), and taking different approaches. Articles are included after an innovative search method, including Dense Query Extraction and automatic cross-referencing. Discussed are the types of groups and emotion models considered in automatic emotion recognition research, common datasets for all modalities, general approaches taken, and reported performances. These performances are discussed, followed by an analysis of the application possibilities of the discussed methods. To ensure clear, replicable, and comparable studies, we suggest research should test on multiple, common datasets and report on multiple metrics, when possible. Implementation details and code should be made available where possible. An area of interest for future work is to build systems with more real-world application possibilities, coping with changing group sizes, different emotional subgroups, and changing emotions over time, while having a higher robustness and working with datasets with reduced biases.

In this paper shows check the traditional, CPWM (computerized pulse width modulation) regulator combine for control of speed by the automatic group the Brushless dc (BLDC) motor, from these strategies, pursue from CPWM speed control system, the main distinction is in the regulator part. Unique drive framework comprises of a steady Proportional Integrator controller and different has an advanced controller. This paper remembers conversation for the execution of both the strategies and their exhibitions are thought about.

Background This work aims at clinically validating a graphical tool developed for treatment plan assessment, named SPIDERplan, by comparing the plan choices based on its scoring with the radiation oncologists (RO) clinical preferences. Methods SPIDERplan validation was performed for nasopharynx pathology in two steps. In the first step, three ROs from three Portuguese radiotherapy departments were asked to blindly evaluate and rank the dose distributions of twenty pairs of treatment plans. For plan ranking, the best plan from each pair was selected. For plan evaluation, the qualitative classification of ‘Good’, ‘Admissible with minor deviations’ and ‘Not Admissible’ were assigned to each plan. In the second step, SPIDERplan was applied to the same twenty patient cases. The tool was configured for two sets of structures groups: the local clinical set and the groups of structures suggested in international guidelines for nasopharynx cancer. Group weights, quantifying the importance of each group and incorporated in SPIDERplan, were defined according to RO clinical preferences and determined automatically by applying a mixed linear programming model for implicit elicitation of preferences. Intra- and inter-rater ROs plan selection and evaluation were assessed using Brennan-Prediger kappa coefficient. Results Two-thirds of the plans were qualitatively evaluated by the ROs as ‘Good’. Concerning intra- and inter-rater variabilities of plan selection, fair agreements were obtained for most of the ROs. For plan evaluation, substantial agreements were verified in most cases. The choice of the best plan made by SPIDERplan was identical for all sets of groups and, in most cases, agreed with RO plan selection. Differences between RO choice and SPIDERplan analysis only occurred in cases for which the score differences between the plans was very low. A score difference threshold of 0.005 was defined as the value below which two plans are considered of equivalent quality. Conclusion Generally, SPIDERplan response successfully reproduced the ROs plan selection. SPIDERplan assessment performance can represent clinical preferences based either on manual or automatic group weight assignment. For nasopharynx cases, SPIDERplan was robust in terms of the definitions of structure groups, being able to support different configurations without losing accuracy.

Cross-Classroom Collaborative Project-Based Learning (C3PjBL) requires the formation of project-groups by pairing student-groups across classrooms. Unfortunately, due to the configuration of these groups, the group formation techniques found in the literature are unable to automatically create project-groups for C3PjBL. This paper describes an automatic project-group formation technique for C3PjBL which utilizes clustering to create homogeneous student-groups, based on the students’ perceived technological and higher-order thinking skills (student characteristics). Student-groups, from different classrooms, are then paired using an optimization technique to form project-groups. In our results, we present a comparison of the performance of a random group formation technique and our technique. We observed that automatic group formation using an n-dimensional space of student characteristics and k-means clustering is more effective than random group formation and, the strategy of forming homogeneous student-groups and heterogeneous project-group for C3PjBL creates more compatible group compositions than random grouping.

Group recommender systems are specialized in suggesting preferable products or services to a group of users rather than an individual by aggregating personal preferences of group members. In such expert systems, the initial task is to identify groups of similar users via clustering approaches as user groups are usually not predefined. However, clustering users into groups commonly suffer from sparsity, scalability, and complexity problems as the content in the domain proliferate. Moreover, group homogeneity and size are the critical parameters for organizing group members and enhancing their satisfaction. In this study, we propose novel automatic user grouping approaches by constructing a binary decision tree via bisecting k -means clustering for enhanced group formation and group size restriction. Furthermore, we propose adopting a genre-based mapping of user ratings into a tiny and dense vector to represent users, which both improves computation time for constructing the binary decision tree and enables eliminating adverse effects of sparsity. Finally, since the quality of group formation is not only dependent on conforming preferences but also to the demographic harmony among members, we further introduce utilizing similarities based on demographic characteristics along with the genre-based similarities. We propose applying two distinct strategies for small and large groups by decorating the genre-based similarities with demographic properties, which leads to a more homogeneous automatic group formation. Experiments performed on real-world benchmark datasets demonstrate that each proposed method outperforms its traditional rival significantly, and the final proposed method achieves significantly more qualified ranked recommendation lists than the state-of-the-art algorithm.

Collaborative learning in the form of group work is becoming increasingly significant in education since interpersonal skills count in modern society. However, teachers often get overwhelmed by the logistics involved in conducting any group work. Valid support for executing and managing such activities in a timely and informed manner becomes imperative. This research introduces an intelligent system focusing on group formation which consists of a parameter setting module and the group member visualization panel where the results of the created group are shown to the user and can be graded. The system supports teachers by applying algorithms to actual learning log data thereby simplifying the group formation process and saving time for them. A pilot study in a primary school mathematics class proved to have a positive effect on students’ engagement and affections while participating in group activities based on the system-generated groups, thus providing empirical evidence to the practice of Computer-Supported Collaborative Learning (CSCL) systems.

In order to improve the quality of the traditional exam management system test paper generation, joining the genetic algorithm of simulated annealing algorithm is put forward, and study at last on the automatic group volume algorithm update the topic the difficulty coefficient. By analyzing the simulation results, the algorithm can improve the efficiency and quality of the volume.

[Abstract] Background Diffusion MRI is the preferred non-invasive in vivo modality for the study of brain white matter connections. Tractography datasets contain 3D streamlines that can be analyzed to study the main brain white matter tracts. Fiber clustering methods have been used to automatically group similar fibers into clusters. However, due to inter-subject variability and artifacts, the resulting clusters are difficult to process for finding common connections across subjects, specially for superficial white matter. Methods We present an automatic method for labeling of short association bundles on a group of subjects. The method is based on an intra-subject fiber clustering that generates compact fiber clusters. Posteriorly, the clusters are labeled based on the cortical connectivity of the fibers, taking as reference the Desikan–Killiany atlas, and named according to their relative position along one axis. Finally, two different strategies were applied and compared for the labeling of inter-subject bundles: a matching with the Hungarian algorithm, and a well-known fiber clustering algorithm, called QuickBundles. Results Individual labeling was executed over four subjects, with an execution time of 3.6 min. An inspection of individual labeling based on a distance measure showed good correspondence among the four tested subjects. Two inter-subject labeling were successfully implemented and applied to 20 subjects and compared using a set of distance thresholds, ranging from a conservative value of 10 mm to a moderate value of 21 mm. Hungarian algorithm led to a high correspondence, but low reproducibility for all the thresholds, with 96 s of execution time. QuickBundles led to better correspondence, reproducibility and short execution time of 9 s. Hence, the whole processing for the inter-subject labeling over 20 subjects takes 1.17 h. Conclusion We implemented a method for the automatic labeling of short bundles in individuals, based on an intra-subject clustering and the connectivity of the clusters with the cortex. The labels provide useful information for the visualization and analysis of individual connections, which is very difficult without any additional information. Furthermore, we provide two fast inter-subject bundle labeling methods. The obtained clusters could be used for performing manual or automatic connectivity analysis in individuals or across subjects. Anillo de Investigación en Ciencia y Tecnología (Chile); 2016-21160342 Fondo Nacional de Desarrollo Científico y Tecnológico; 1190701 Anillo de Investigación en Ciencia y Tecnología (Chile); ACT172121 Anillo de Investigación en Ciencia y Tecnología (Chile); FB0008

Purpose This study compared the technical parameters and clinical outcomes of manual and automatic image fusion techniques of ultrasonography and magnetic resonance imaging/computed tomography for radiofrequency ablation (RFA) of hepatic tumors. Methods Seventy consecutive patients (male:female=47:23, 67.1±10.9 years old) who underwent RFA for hepatic tumors were prospectively enrolled and randomly assigned to the manual or automatic registration group. Two operators performed RFA with one of two imaging fusion techniques. Technical parameters (the registration error, time required for image registration, number of point registrations) and clinical outcomes (technical success, technical effectiveness, local tumor progression [LTP]-free survival, and progression-free survival [PFS]) were compared. Results The automatic group contained 35 patients with hepatocellular carcinoma, while the manual group included 34 hepatocellular carcinoma patients and a patient with colon cancer liver metastasis. The registration error, time required for registration, and number of point registrations were 5.7±4.3 mm, 147.8±78.2 seconds, and 3.26±1.20 in the automatic group, and 6.3±5.0 mm, 150.3±89.7 seconds, and 3.20±1.13 in the manual group, respectively. The technical success and effectiveness rates were both 97.1% in the automatic group and both 100.0% in the manual group. The above differences were not significant. The LTP-free survival and PFS (28.3 and 21.2 months in the automatic group, and 29.0 and 24.9 months in the manual group, respectively) showed no significant between-group differences during a median 20.1-month follow-up period. Conclusion The technical parameters and clinical outcomes of automatic image fusion were not significantly different from those of manual image fusion for RFA of hepatic tumors.

This article proposes an automatic group formation method applying the particle swarm optimization (PSO) algorithm to boost the quality of students' online interactions. The groups were heterogeneous regarding their levels of knowledge and their interests, and three different leadership roles were distributed among group members. A case study with 66 undergraduate students was performed. Discourse analysis was applied using two coding schemes to measure the critical thinking apparent in the students' online discussions and evaluate the socio-cognitive aspects of group interactions. The results provided evidence that groups of undergraduate students formed by the proposed method achieved better scores in most categories analyzed when compared to the randomly formed groups.

Despite group formation in learning environments is commonly and successfully approached, there is a gap in the research literature with respect to its application in corporative learning. Regarding that creativity is as an important factor to increase innovation in companies, in the present research, we propose a group formation method, considering preferred roles and functional diversity, aiming to improve creativity in collaborative learning at workplace. We employed Tabu Search algorithm to automatically form groups based on Nonaka's knowledge creation theory and preferred roles from Belbin’s model. We performed a case study to compare the quality of socio-cognitive interactions duringcollaborative learning in groups formed by the proposed method and randomly formed groups. The results show that groups formed by preferred roles and functional diversity are more creative and present enhanced fluency and more elaborated products in comparison to randomly formed groups.

There has been a tremendous advance in machine learning techniques especially for automatic group activity recognition (GAR) over the past few decades. This review article surveys the modern advancement made towards video-based group activity recognition technique. Various applications, including video surveillance systems, sports analytics and human behaviour for robotics characterization, require a group activity recognition system. Comprehensive reviews of machine learning (ML) techniques like hidden Markov models (HMMs), graphical method and support vector machines employed in GAR are being discussed. A comprehensive review on the latest progress in deep learning model has delivered important developments in GAR performance; those are also presented. The main purpose of this survey is to broadly categorize and analyse GAR according to handcrafted features based on machine learning model and learned features based on deep model. Various GAR models illustrated by considering activities of individual person, person-to-person interaction, person-to-group interaction and group interaction using temporal sequence information from video frames for recognition of group activity are discussed. The review facilitates in diverse applications, and the models described in different applications present specifically in surveillance, sport analytics, video summary, etc.

Group cohesiveness is a compelling and often studied composition in group dynamics and group performance. The enormous number of web images of groups of people can be used to develop an effective method to detect group cohesiveness. This paper introduces an automatic group cohesiveness prediction method for the 7th Emotion Recognition in the Wild (EmotiW 2019) Grand Challenge in the category of Group-based Cohesion Prediction. The task is to predict the cohesive level for a group of people in images. To tackle this problem, a hybrid network including regression models which are separately trained on face features, skeleton features, and scene features is proposed. Predicted regression values, corresponding to each feature, are fused for the final cohesive intensity. Experimental results demonstrate that the proposed hybrid network is effective and makes promising improvements. A mean squared error (MSE) of 0.444 is achieved on the testing sets which outperforms the baseline MSE of 0.5.

This paper proposes a database for group level emotion recognition in videos. The motivation is coming from the large number of information which the users are sharing online. This gives us the opportunity to use this perceived affect for various tasks. Most of the work in this area has been restricted to controlled environments. In this paper, we explore the group level emotion and cohesion in a real-world environment. There are several challenges involved in moving from a controlled environment to real-world scenarios such as face tracking limitations, illumination variations, occlusion and type of gatherings. As an attempt to address these challenges, we propose a ‘Video level Group AFfect (VGAF), database containing 1,004 videos downloaded from the web. The collected videos have a large variations in terms of gender, ethnicity, the type of social event, number of people, pose, etc. We have labelled our database for group level emotion and cohesion tasks and proposed a baseline based on the Inception V3 network on the database.

In this paper, an approach based on genetic algorithms is proposed to form groups in collaborative learning scenarios, considering the students’ personality traits as a criterion for grouping. This formation is carried out in two stages: In the first, the information of the students is collected from a psychometric instrument based on the Big Five personality model, whereas, in the second, this information feeds a genetic algorithm that is in charge of performing the grouping iteratively, seeking for an optimal formation. The results presented here correspond to the functional and empirical validation of the approach. It is found that the described methodology is useful to obtain groups with the desired characteristics. The specific objective is to provide a strategy that makes it possible to subsequently assess in the context what type of approach (homogeneous, heterogeneous, or mixed) is the most appropriate to organize the groups.

Due to the rapid growth of modern mobile devices, users can capture a variety of videos at anytime and anywhere. The explosive growth of mobile videos brings about the difficulty and challenge on categorization and management. In this paper, we propose a novel approach to annotate group activities for mobile videos, which helps tag each person with an activity label, thus helping users efficiently manage the uploaded videos. To extract rich context information, we jointly model three co-existing cues including the activity duration time, individual action feature and the context information shared between person interactions. Then these appearances and context cues are modeled with a structure learning framework, which can be solved by inference with a greedy forward search. Moreover, we can infer group activity labels of all the persons together with their activity durations, especially for the situation with multiple group activities co-existing. Experimental results on mobile video dataset show that the proposed approach achieves outstanding results for group activity classification and annotation.

Quasi-automatic semigroups are extensions of a Cayley graph of an automatic group. Of course, a quasi-automatic semigroup generalizes an automatic semigroup. We observe that a semigroup [Formula: see text] may be automatic only when [Formula: see text] is finitely generated, while a semigroup may be quasi-automatic but it is not necessary finitely generated. Similar to the usual automatic semigroups, a quasi-automatic semigroup is closed under direct and free products. Furthermore, a semigroup [Formula: see text] is graph automatic if and only if [Formula: see text] with a zero element adjoined is graph automatic, and also a semigroup [Formula: see text] is graph automatic if and only if [Formula: see text] with an identity element adjoined is graph automatic. However, the class of quasi-automatic semigroups is a much wider class than the class of automatic semigroups. In this paper, we show that every automatic semigroup is quasi-automatic but the converse statement is not true (see Example ??). In addition, we notice that the quasi-automatic semigroups are invariant under the changing of generators, while a semigroup may be automatic with respect to a finite generating set but not the other. Finally, the connection between the quasi-automaticity of two semigroups [Formula: see text] and [Formula: see text], where [Formula: see text] is a subsemigroup with finite Rees index in [Formula: see text] will be investigated and considered.

Creative thinking comprises two main components: divergent and convergent thinking. The Remote Associates Test (RAT) was designed to examine the ability to form associative elements into new combinations, however it is widely used as a general creativity measure, without sub-dividing it to its components. Our goal here was to explore the sub-components of the RAT, aiming to link them to the angular gyrus (AG) activation. The AG seems as a good candidate to host both aspects of the RAT, as neuroimaging studies observed deactivation in the AG while participants were engaged in creative tasks, however it also seems to play a role in arithmetic solution retrieval and automatic knowledge retrieval. Our objective was therefore to test whether transcranial direct current stimulation (tDCS) of the AG will influence creative and automatic performance in the RAT. In Experiment 1, in the creative group we administrated cathodal right AG stimulation in order to deactivate the AG aiming to improve divergent features of the RAT. In the automatic group, we administrated double anodal AG stimulation in order to improve convergent features of the RAT and included a control SHAM condition in each group. Experiment 2 activated the AG by stimulation and arithmetic training. We hypothesized that anodal stimulation of the AG will improve automatic convergent features as in Experiment 1. In addition, we expected the arithmetic training to improve the automatic score in the RAT. In Experiment 1, activation of the AG interrupted creative abilities and enhanced automatic abilities. Additionally, deactivation of the AG enhanced creative abilities reflecting divergent thinking. While in Experiment 2 both stimulation and arithmetic training resulted in higher automatic performance. We argue that the RAT measures automatic rather than creative abilities.

This paper proposes a pipeline for automatic group-level affect analysis. A deep neural network-based approach, which leverages on the facial-expression information, scene information and a high-level facial visual attribute information is proposed. A capsule network-based architecture is used to predict the facial expression. Transfer learning is used on Inception-V3 to extract global image-based features which contain scene information. Another network is trained for inferring the facial attributes of the group members. Further, these attributes are pooled at a group-level to train a network for inferring the group-level affect. The facial attribute prediction network, although is simple yet, is effective and generates result comparable to the state-of-the-art methods. Later, model integration is performed from the three channels. The experiments show the effectiveness of the proposed techniques on three ‘in the wild’ databases: Group Affect Database, HAPPEI and UCLA-Protest database.

With the accessibility to information, users often face the problem of selecting one item (a product or a service) from a huge search space. This problem is known as information overload. Therefore to help them in searching items on internet we propose a recommender system which recommends items on social networking site considering group members opinion. An important issue for the RS that has greatly captured the attention of researchers is the new user cold-star problem, which occurs when there is a new user that has been registered to the system and no prior rating of this user is found in the rating table. In this paper, items will be recommended to the group of user having similar interest and liking. Thus, the proposed system implements the advanced recommendation model in which users will satisfy in multiple ways. Recommendation will be done in two ways, individual recommendation and group recommendation. Individual recommendation contain preference wise recommendation and group recommendation contain similar profile wise and social behavior wise recommendation. So, both individual as well as group recommendation will be there. So for group recommendation, we propose two automatic group detection method i.e. similar profile wise and social behavior wise. Social behavior wise group detection method is used to avoid the cold-start problem, which also reduce the overhead as well as increase the accuracy. The experimental results indicate that social behavior wise group detection method achieves better accuracy and computation time than the relevant methods. In any case user will get the recommendation as per his likings.

Automatic group delay tuning of a lX4 RF Si-photonic asymmetric binary tree beamforming network structure is achieved with ring resonator drop port photodetectors providing electrical monitor information. The proposed tuning approach calibrates the group delay of an entire optical beamforming network (OBFN) that operates at 30GHz with 2GHz bandwidth. Compensation of process variations and thermal crosstalk is achieved with the 4 calibrated optical outputs displaying equal 41ps delay intervals.

The recent advancement of social media has given users a platform to socially engage and interact with a larger population. Millions of images and videos are being uploaded everyday by users on the web from different events and social gatherings. There is an increasing interest in designing systems capable of understanding human manifestations of emotional attributes and affective displays. As images and videos from social events generally contain multiple subjects, it is an essential step to study these groups of people. In this paper, we study the problem of happiness intensity analysis of a group of people in an image using facial expression analysis. A user perception study is conducted to understand various attributes, which affect a person’s perception of the happiness intensity of a group. We identify the challenges in developing an automatic mood analysis system and propose three models based on the attributes in the study. An ‘in the wild’ image-based database is collected. To validate the methods, both quantitative and qualitative experiments are performed and applied to the problem of shot selection, event summarisation and album creation. The experiments show that the global and local attributes defined in the paper provide useful information for theme expression analysis, with results close to human perception results.

We show that the higher rank lamplighter groups, or Diestel–Leader groups Γ d ( q ) for d ≥ 3 , are graph automatic. This introduces a new family of graph automatic groups which are not automatic.

We consider user to cell association in a heterogeneous network with a mix of LTE/3G and WiFi cells. Individual user preferences are often neglected when a user to cell association decision is made. In this paper we propose use of a recommender system to inform the mapping of users to cells. We demonstrate the effectiveness of the proposed grouped-based user to cell associations for a set of synthetically generated user/cell ratings.

The growth function is the generating function for sizes of spheres around the identity in Cayley graphs of groups. We present a novel method to calculate growth functions for automatic groups with normal form recognizing automata that recognize a single normal form for each group element, and are at most context free in complexity: context free grammars can be translated into algebraic systems of equations, whose solutions represent generating functions of their corresponding non-terminal symbols. This approach allows us to seamlessly introduce weightings on the growth function: assign different or even distinct weights to each of the generators in an underlying presentation, such that this weighting is reflected in the growth function. We recover known growth functions for small braid groups, and calculate growth functions that weight each generator in an automatic presentation of the braid groups according to their lengths in braid generators., Comment: 7 pages

Detecting groups plays an important role for group activity detection. In this paper, we propose an automatic group activity detection by segmenting the video sequences automatically into dynamic clips. As the first step, groups are detected by adopting a bottom-up hierarchical clustering, where the number of groups is not provided beforehand. Then, groups are tracked over time to generate consistent trajectories. Furthermore, the Granger causality is used to compute the mutual effect between objects based on motion and appearances features. Finally, the Hierarchical Dirichlet Process is used to cluster the groups. Our approach not only detects the activity among the objects of a particular group (intra-group) but also extracts the activities among multiple groups (inter-group). The experiments on public datasets demonstrate the effectiveness of the proposed method. Although our approach is completely unsupervised, we achieved results with a clustering accuracy of up to 79.35\(\%\) and up to 81.94\(\%\) on the Behave and the NUS-HGA datasets.

Although there is significant evidence regarding benefits of small group collaboration in small-scale contexts, several challenges have been detected about the use of collaborative learning in MOOCs. Group formation, which is a crucial activity in order to achieve effective collaboration, is scarcely covered in MOOC platforms, which do not allow the formation of teams using criteria defined by the instructors. This paper presents an exploratory study conducted in a seven-week MOOC, comparing our group formation proposal, based on students’ activity criteria, to a baseline grouping function provided by the platform. We analyse the impact of each grouping approach over group performance, group activity, and student satisfaction. The results show initial evidence about the advantages of using the criteria-based group formation approach regarding student satisfaction and group interactions.

We consider the kernel of the natural projection from the Artin group of dihedral type I2(k) to the corresponding Coxeter group, which we call the pure Artin group of dihedral type. We present a rational function expression for the spherical growth series of the pure Artin group of dihedral type with respect to a natural generating set.

Human activity recognition is important for healthcare and lifestyle evaluation. In this paper, a novel method for activity recognition by jointly considering motion sensor data recorded by wearable smart watches and image data captured by RGB-Depth (RGB-D) cameras is presented. A normalized cross correlation based mapping method is implemented to establish association between motion sensor data with corresponding image data from the same person in multi-person situations. Further, to improve the performance and accuracy of recognition, a hierarchical structure embedded with an automatic group selection method is proposed. Through this method, if the number of activities to be classified is changed, the structure will be changed correspondingly without interaction. Our comparative experiments against the single data source and single layer methods have shown that our method is more accurate and robust.

In this research study, we propose an automatic group activity recognition approach by modelling the interdependencies of group activity features over time. Unlike in simple human activity recognition approaches, the distinguishing characteristics of group activities are often determined by how the movement of people are influenced by one another. We propose to model the group interdependences in both motion and location spaces. These spaces are extended to time-space and time-movement spaces and modelled using Kernel Density Estimation (KDE). Such representations are then fed into a machine learning classifier which identifies the group activity. Unlike other approaches to group activity recognition, we do not rely on the manual annotation of pedestrian tracks from the video sequence.

We study the geometry of abelian subgroups relative to combings (normal forms) of automatic groups and groups which act properly and cocompactly on spaces of nonpositive curvature. We prove that in general combings which arise in the latter context are incompatible (in a sense which we make precise) with regular combings, the fundamental objects of automatic group theory.

If G is a finitely generated group that is abelian or word-hyperbolic and H is an asynchronously combable group then every split extension of G by H is asynchronously combable. The fundamental group of any compact 3-manifold that satisfies the geometrization conjecture is asynchronously combable. Every split extension of a word-hyperbolic group by an asynchronously automatic group is asynchronously automatic. © 1993 Birkhäuser Verlag.

One goal of collaborative learning is to maximize the learning performance of all participating students. In order to achieve this aim, the first step is to consider how to assist instructors in forming well-structured collaborative learning groups with a good work atmosphere to promote successful outcomes for all members. Generally, understanding levels and interests of students are two grouping criteria that are usually considered by practicing instructors. Nevertheless, when the instructors face a large number of students, simultaneously considering the two grouping criteria to form the students in an appropriate collaborative learning context is almost impossible. To address this problem, this study formulates a group composition problem to model the formation of collaborative learning groups that satisfy the two grouping criteria. Moreover, this study is based on a novel approach called particle swarm optimization (PSO) to propose an enhanced PSO (EPSO) for composing well-structured collaborative learning groups. In addition, the experimental results have demonstrated that the proposed approach is an applicable and robust method that can aid instructors in planning different kinds of collaborative learning processes.