Cumulative Learning Research Articles

A hybrid deep model with cumulative learning for few-shot learning

A hybrid deep model with cumulative learning for few-shot learning

The conceptual nuances of technology-supported learning in engineering

ABSTRACT Enabling theory-practice bridging in engineering education is essential for developing twenty-first century graduate capabilities. Massification, resource constraints, and technological development have resulted in significant shifts to alternative forms of practical engagement, such as the use of online laboratories, but how do these contribute to learning? Based on three illustrative case studies at a research-intensive institution in the Global South, this paper offers a conceptualisation of the degrees of complexity entailed in multimodal approaches to teaching Fluid Mechanics, Finite Element Analysis and Control Systems at different stages of their respective programmes The paper examines the different levels of abstract-concrete learning when students engage with verbal, symbolic, graphic and physical representational artefacts designed to enable cumulative learning. The conceptual instruments are theoretically and methodologically drawn from Legitimation Code Theory dimensions, which lend themselves to the graphic analysis of knowledge practices. It is suggested that the explicit integration of and shifting between levels of abstraction and complexity with different kinds of technologies enables the kind of cumulative learning necessary to prepare technically-equipped graduates for complex twenty-first century engineering contexts.

A cluster's internationalization as a catalyst for its innovation system's access to global markets

PurposeThe paper aims to illuminate the platform created by a cluster organization to facilitate its internationalization and thereby enhance its regional innovation system partners' competitiveness by providing access to global value chains and boosting innovativeness.Design/methodology/approachThe study draws upon the interaction approach, focusing on the interaction process, interaction partners, relationship atmosphere, and relationship environment. A qualitative study was conducted at Future Position X, a Swedish cluster organization. A total of 58 interviews were conducted, including 48 face-to-face in-depth interviews between 2017 and 2019 with six key informants at FPX, representatives from 28 SMEs, ten members of regional innovation systems to which FPX belongs, and four process leaders of regional and local networks, in addition to online interviews with ten members of the regional innovation systems conducted via Microsoft Teams in March 2021. The time span of the study provides a longitudinal perspective.FindingsThe FPX cluster collaborates with actors in the quadruple helix, maintaining a mindset that has led to a number of new partner agreements in the global arena to secure the resources and expertise necessary for cluster activities, and thereby ensuring firms in FPX networks access to platforms for international expansion. Internationalization thus expands the cluster's knowledge base beyond the traditional environment of its member firms.Research limitations/implicationsVery few innovations arise from the isolated work of a lone genius. Instead, most innovation is achieved through complex, interactive, iterative and cumulative learning processes in which a variety of actors are involved. The FPX cluster organization's internationalization platform is therefore vital to the internationalization of its partners since cluster actors lack the time, resources, knowledge, experience, and networks required to break into international markets singlehandedly.Practical implicationsThis study suggests that, for practitioners and researchers alike, the growing importance and relevance of the regional innovation system cannot be overemphasized. It also holds policy and societal implications in that FPX's global network helps regional SMEs to internationalize, in addition to inspiring international firms to establish operations in the Gävleborg region, thereby helping to strengthen the overall GIS environment. Internationalization also expands the FPX cluster's knowledge base beyond the traditional environment of its firms, an example of this being the construction start of a Microsoft data centre in the region in 2020.Social implicationsFPX is financed through taxation and grant funding. By initiating projects, creating relationships and building collaborations, FPX thus contributes to collaboration between business, academia and the public sector. FPX also contributes to knowledge development of new technology by creating meeting places and networks around digital issues, such as GIS, AI, the IoT and blockchain technology.Originality/valueWhile earlier research has concentrated on endogenous gaps critical to cluster dynamics, comparatively little attention has been paid to exogenous gaps, i.e. linkages between regional clusters and innovation partners elsewhere in the world. This study showcases the richness of interactions in the cluster against the background of wider, global innovation interactions. Future research should examine other vital questions that remain unanswered, e.g. by measuring and exploring the extent to which regional innovation systems can contribute to long-term economic growth for society.

Cumulative dual-branch network framework for long-tailed multi-class classification

The long-tailed data distribution problem (i.e., a few classes account for majority data, while most classes account for minority data) is widespread in large-scale and real-world datasets, and it poses a huge challenge to the computer vision field. Existing methods of long-tailed classification mainly focus on re-sampling, re-weighting, and transfer learning. Although class imbalance learning can yield better long-tailed classification performance, the feature representative ability of the feature extraction network is damaged to a certain extent. To deal with these issues, the present work proposes a novel cumulative dual-branch network framework (CDBNF), which takes into account the class imbalance learning and feature representation learning at the same time by the dual-branch network architecture. In CDBNF, the class imbalance learning branch greatly improves the classification performance of tail classes, while the few-shot learning branch enhances the feature representative ability. Furthermore, a cumulative learning strategy (CLS) is proposed in CDBNF to make it pay more attention to the tail classes gradually in the training process. The effectiveness and practicability of the proposed CDBNF are verified by the four benchmark datasets. Experimental results show that the classification performance of CDBNF is superior to other state-of-the-art methods, while the intra-class feature variance is smaller than other state-of-the-art methods.

Do learning style preferences influence the cumulative gross point average and self directed learning hours in dental students: a preliminary study

BackgroundLearning styles influence the outcome of the student performances based on preliminary data available. To evaluate whether the learning styles discriminate the cumulative gross point average (CGPA) scores and self-directed learning hours (SDL) in an integrated curriculum of dental students.MethodsParticipants in this blinded cross-sectional study were dental students enrolled in the Bachelor of Dental & Oral Surgery program at XXXX College of Dentistry. An online survey (Kolb Learning Style Inventory) was used to collect data. It has four sections: Concrete Experience (CE), Conceptualization (AC), Reflective Observation (RO), and Active Experimentation (AE). Questionnaire was distributed electronically to students of Academic level 1 to 5, selected by using non-probability quota sampling technique. In addition to learning style assessment the questionnaire also included measures to obtain data such as gender, academic level, CGPA score, and SDL hours of participants. The CGPA scores were categorized into average (3 to 3.6), good (3.7 to 4.2), excellent (4.3 to 4.7) and outstanding (> 4.7) as well as SDL in to three classes as, < 1 h, > 1 but < 3 h and > 3 h. Discriminant function test was computed to assess the effectiveness of discrimination by the learning styles in GPA and SDL.ResultsThe study's questionnaire was completed by 198 participants (43% females and 57% males). Learning styles were discriminated by excellent category of CGPA scores that presented 72.1% group membership whereas in case of outstanding category presented the least as 17% group membership. Learning styles were discriminated by > 2 but < 3 h category of SDL hours that presented 69.7% group membership.ConclusionLearning styles can be used to discriminate the student academic performances and self-directed learning hours. Among the different category of CGPA participants with outstanding performance represent a good prediction for learning styles preferences. Participants with varying SDL hours also influenced the learning styles.

Neurosymbolic Systems of Perception and Cognition: The Role of Attention.

A cognitive architecture aimed at cumulative learning must provide the necessary information and control structures to allow agents to learn incrementally and autonomously from their experience. This involves managing an agent's goals as well as continuously relating sensory information to these in its perception-cognition information processing stack. The more varied the environment of a learning agent is, the more general and flexible must be these mechanisms to handle a wider variety of relevant patterns, tasks, and goal structures. While many researchers agree that information at different levels of abstraction likely differs in its makeup and structure and processing mechanisms, agreement on the particulars of such differences is not generally shared in the research community. A dual processing architecture (often referred to as System-1 and System-2) has been proposed as a model of cognitive processing, and they are often considered as responsible for low- and high-level information, respectively. We posit that cognition is not binary in this way and that knowledge at any level of abstraction involves what we refer to as neurosymbolic information, meaning that data at both high and low levels must contain both symbolic and subsymbolic information. Further, we argue that the main differentiating factor between the processing of high and low levels of data abstraction can be largely attributed to the nature of the involved attention mechanisms. We describe the key arguments behind this view and review relevant evidence from the literature.

Single Model Deep Learning on Imbalanced Small Datasets for Skin Lesion Classification.

Deep convolutional neural network (DCNN) models have been widely explored for skin disease diagnosis and some of them have achieved the diagnostic outcomes comparable or even superior to those of dermatologists. However, broad implementation of DCNN in skin disease detection is hindered by small size and data imbalance of the publically accessible skin lesion datasets. This paper proposes a novel single-model based strategy for classification of skin lesions on small and imbalanced datasets. First, various DCNNs are trained on different small and imbalanced datasets to verify that the models with moderate complexity outperform the larger models. Second, regularization DropOut and DropBlock are added to reduce overfitting and a Modified RandAugment augmentation strategy is proposed to deal with the defects of sample underrepresentation in the small dataset. Finally, a novel Multi-Weighted New Loss (MWNL) function and an end-to-end cumulative learning strategy (CLS) are introduced to overcome the challenge of uneven sample size and classification difficulty and to reduce the impact of abnormal samples on training. By combining Modified RandAugment, MWNL and CLS, our single DCNN model method achieved the classification accuracy comparable or superior to those of multiple ensembling models on different dermoscopic image datasets. Our study shows that this method is able to achieve a high classification performance at a low cost of computational resources and inference time, potentially suitable to implement in mobile devices for automated screening of skin lesions and many other malignancies in low resource settings.

Exploring Determinants of Desirable Financial Behaviors Using Decision Tree Analysis Evidence From Four Waves of National Financial Capability Study

The purpose of this article is to utilize decision tree (DT) analysis to examine the relationship between income level, financial satisfaction, financial confidence, financial knowledge, and several demographics with a goal of better understanding desirable financial behavior. The emphasis of this analysis is focused particularly upon better understanding the role of financial knowledge in desirable behavior outcomes. DT analysis is most useful when an analysis includes numerous variables and solving problems where the cumulative learning process is inherent. Our DT analysis of four FINRA National Financial Capability datasets (2009, 2012, 2015, and 2018) suggest that financial knowledge is a relevant variable only under specific circumstances and for respondents with relatively higher income levels. Key variables in the DT analysis included income level and financial satisfaction.

PF2RM: A Power Fault Retrieval and Recommendation Model Based on Knowledge Graph

Energy saving and emission reduction have become common concerns in countries around the world. In China, with the implementation of the new strategy of “carbon peak and neutrality” and the rapid development of the new smart grid infrastructure, the amount of data of actual power grid dispatching and fault analysis show exponential growth, which has led to phenomena such as poor supervision effectiveness and difficulty in handling faults in the process of grid operation and maintenance. Existing research on retrieval recommendation methods has had a lower accuracy rate at cold-start due to a small sample of user interactions. In addition, the cumulative learning of user personalization during general retrieval results in a poor perception of potential interest. By constructing a power knowledge graph, this paper presents a power fault retrieval and recommendation model (PF2RM) based on user-polymorphic perception. This model includes two methods: the power fault retrieval method (PFR) and the user-polymorphic retrieval recommendation method (UPRR). First, we take the power grid fault dispatching business as the core and reconstruct the ontology layer of the power knowledge graph. The PFR method is used to design the graph-neighbor fault entity cluster to enhance the polymerization degree of a fault implementation scenario. This method can solve the search cold-start recommendation problem. At the same time, the UPRR method aims to form user retrieval subgraphs of the past-state and current-state and make a feature matching for the graph-neighbor fault entity cluster, and then realize the accurate prediction of the user’s general search intention. The model is compared with other current classical models through the evaluation of multiple recommendation evaluation metrics, and the experimental results show that the model has a 3–8% improvement in the cold-start recommendation effect and 2–10% improvement in regular retrieval. The model has the best average recommendation performance in multiple metrics and has good results in fault analysis and retrieval recommendation. It plays a helpful role in intelligent operation and maintenance of the power grid and auxiliary decision-making, and effectively improves the reliability of the power grid.

MCLA: Research on cumulative learning of Markov Logic Network

Compared with other probabilistic semantic learning algorithms, Markov Logic Network (MLN) learning can integrate existing knowledge fragments. However, the knowledge fragments increase as the MLN learning algorithm traverses candidate objects, making the execution time of the MLN learning algorithm long. To shorten the execution time, an MLN Cumulative Learning Algorithm (MCLA) that combines the MLN learning algorithm and cumulative learning is proposed. MCLA can unify the existing knowledge and the new–old knowledge, perform multi-task learning under appropriate circumstances, and directly call the existing knowledge to learn the new knowledge. This paper applies MCLA to single-resident indoor activity scenarios to verify the effectiveness of the learning algorithm. Experiments have proved that MCLA not only ensures the accuracy and improves the versatility of knowledge, but also dramatically decreases the learning time and effectively manages the knowledge memory of the MLN.

Characterisation of Deficits and Sex Differences in Verbal and Visual Memory/Learning in Bipolar Disorder.

Cognitive impairment is consistently reported in bipolar disorder (BD), but few studies have characterised which memory component processes are affected. Further, it is unknown whether the component processes underlying memory impairment are moderated by sex. The present study examined diagnosis and sex differences in both verbal and visual memory/learning domains in patients with BD and psychiatrically healthy controls. Verbal and visual memory/learning were measured using the Hopkins Verbal Learning Test-Revised (HVLT-R) and Brief Visuospatial Memory Test-Revised (BVMT-R). 114 patients with BD (n = 50 males, n = 64 females), were compared to 105 psychiatrically healthy controls (n = 42 males, n = 63 females). Patients with BD had worse performance in verbal and visual immediate and total recall, verbal and visual delayed free recall, and verbal recognition discrimination scores, but there were no group differences in learning slopes and cumulative learning index scores. There were trends for BD females to outperform BD males in visual memory/learning free recall and cumulative learning, but these results did not survive multiple testing correction. These findings did not change in a secondary sensitivity analysis comparing only strictly euthymic BD patients to controls (n = 64). The present study found trait-like verbal and visual memory/learning impairment in BD that was attributable to deficient encoding and/or consolidation processes rather than deficits in learning. We did not find marked sex differences in either visual or verbal memory/learning measures, although some trend level effects were apparent and deserve exploration in future studies.

Enabling cumulative learning in user-oriented research for root, tuber and banana crop breeding

Summary User-oriented research is important in breeding improved genotypes, for developing and validating product concepts (mostly involving trait prioritisation), as well as evaluating breeding products in use situations (mostly involving participatory variety evaluation). This paper examines key aspects that enable cumulative learning in user-oriented research for root, tuber and banana (RTB) crop breeding. We reviewed empirical user-oriented studies on RTB crops published between 1996 and 2020. We examined the ability of user-oriented research to foster cumulative learning by examining four key aspects: spatial and temporal coverage; gender aspects; the range of traits considered and publishing practices as evident in reports and datasets. We conclude that user-oriented studies have received attention in RTB breeding but fall short of enabling cumulative learning. Substantial investments in methodology development and capacity are needed to bring greater coherence to this field and enable cumulative learning about user perspectives to iteratively increase the fit between improved genotypes and user preferences.

Supplemental Material for Executive Function Mediates the Association Between Cumulative Risk and Learning in Ghanaian Schoolchildren

Supplemental Material for Executive Function Mediates the Association Between Cumulative Risk and Learning in Ghanaian Schoolchildren

Higher Order Thinking by Setting and Debriefing Tasks in Dutch Geography Lessons.

Tasks are a powerful instrument for geography teachers, as they let students engage with the subject. To advance the cumulative learning of students, teachers have to make sure that students learn how to deal with complex and abstract knowledge structures. In the Netherlands, teachers face a dilemma when it comes to task setting: the intended curriculum aims for a considerable part at (parts of) higher order thinking, whereas the high-stakes exams have a clear focus on the use of thinking strategies. This paper explores the task setting and debriefing of Dutch geography teachers by analyzing twenty-three videotaped lessons in upper secondary education by using the Geography Task Categorization Framework. The results show that Dutch teachers mostly rely on textbooks when setting tasks. The focus lies on reproduction and the use of thinking strategies. Tasks aiming at (parts of) higher order thinking are barely used. Furthermore, teachers use tasks from previous high-stakes exams already used in an early stage of upper secondary education. In the debriefing of tasks, teachers move from simple and concrete to complex and abstract knowledge and vice versa. However, most of these movements aim at simplifying knowledge structures. In the observed lessons, curriculum aims at the level of (parts of) higher order thinking are not achieved. The evaluative rules as set by the high-stakes exams and the type of tasks offered by textbooks seem to be dominant.

Clinical and radiological outcomes of robotic-assisted unicompartmental knee arthroplasty: Early lessons from the first 100 consecutive knees in 85 patients

IntroductionRobotic-assisted unicompartmental knee arthroplasty (UKA) is associated with improved component positioning and comparable short- and mid-term implant survivorship with manual UKA. This study aims to evaluate clinical and radiological outcomes following robotic-assisted UKA as well as any potential learning-curves associated with the introduction of such new technology. MethodsProspective study of patients undergoing robotic-assisted UKA. Outcome measures were patient-reported outcome measures (PROMs) including Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC), Knee Society Score (KSS) and Oxford Knee Score (OKS), complications, implant survivorship, component positioning and learning curve. ResultsEighty-five patients comprising 100 knees were recruited and followed up for 21.0 ± 4.3 months. At two years, there were significant and sustained improvements in PROMs and 100% implant survivorship rate. A high degree of implant accuracy was achieved with the robotic system. A cumulative learning curve of 20 cases was noted. ConclusionRobotic-assisted UKA achieves excellent implant accuracy and clinical outcomes in the short-term. Long-term follow up is needed to evaluate this relationship.

Effects of cumulative energy aid projects on renewable energy generation capacity

Development cooperation plays an important role in helping aid recipient countries follow low carbon development pathways. Although the nature of cooperation varies across donors and recipients, intentions of renewable energy technology diffusion and structural changes are common for projects in the energy sector. Aid projects promoting renewable energy in developing countries are seldom assessed through econometric methods as the varieties and differences across projects and countries limit generalization. This paper attempts to explore how cumulative number of renewable energy aid projects enhances generation capacity in recipient countries. Using fixed-effect models on such type of projects from 2000 to 2013, we examine the effects of cumulative energy aid project numbers and other control variables on the yearly changes in generation capacity. A positive and statistically significant relationship was found suggesting cumulative learning effect from implemented projects. Presence of host country policies was also found to be important for growth in capacity. Findings of this study adds to the debate of whether density of energy aid projects leads to donor competition or greater fragmentation. Insights from this study can also improve aid providers’ strategy for interventions in the energy sector of recipient countries.

Persistence and learning effects in design innovation: Evidence from panel data

This paper explores persistence and learning effects in the aesthetic and symbolic dimensions of design innovation. By combining insights from innovation economics and design studies, we discuss design innovation as the result of firm-specific cumulative learning. We then conceptualise design and product innovation as complementary processes whose interplay may lead to learning effects across different dimensions of knowledge creation. We provide quantitative evidence for these insights applying dynamic probit and bivariate probit models to a longitudinal dataset of manufacturing firms based in Spain for the period 2007–2016. Our findings confirm the presence of persistence effects in design innovation, offering novel evidence in support of the view whereby design is an iterative process shaped by the knowledge generated through firms’ previous engagement with design. In addition, the results contribute to our understanding of the role of design beyond its functional dimension, pointing to mutually reinforcing effects between aesthetic and symbolic design and product innovation.

A Seasonal Comparison of the Effectiveness of Parent Engagement on Student Literacy Achievement

Parent engagement is often promoted as a remedy for reducing achievement gaps between students from high socio-economic and low socio-economic backgrounds. However, researchers have found mixed results when examining parent engagement and student outcomes. Drawing on a study investigating the effectiveness of summer literacy camps offered by schools in Ontario, I compare the influence of parent engagement on two outcomes: (1) spring snapshot of cumulative learning, and (2) summer literacy growth/loss. In considering summer learning in regression analysis, I aim to investigate the effect of parent engagement without the influence of schools during the academic year. Out of 14 parent engagement measures, I find only three (parents’ aspirations, home resources, discussions of school with children) are positive predictors of spring literacy outcomes and that none predict summer literacy growth/loss. Family socio-economic status remains a powerful predictor of achievement for both outcomes. I interpret my findings within three proposed mechanisms of parent engagement: cultivation ethic, realist reaction, and expressive logic.

A Seasonal Comparison of the Effectiveness of Parent Engagement on Student Literacy Achievement

Parent engagement is often promoted as a remedy for reducing achievement gaps between students from high socio-economic and low socio-economic backgrounds. However, researchers have found mixed results when examining parent engagement and student outcomes. Drawing on a study investigating the effectiveness of summer literacy camps offered by schools in Ontario, I compare the influence of parent engagement on two outcomes: (1) spring snapshot of cumulative learning, and (2) summer literacy growth/loss. In considering summer learning in regression analysis, I aim to investigate the effect of parent engagement without the influence of schools during the academic year. Out of 14 parent engagement measures, I find only three (parents’ aspirations, home resources, discussions of school with children) are positive predictors of spring literacy outcomes and that none predict summer literacy growth/loss. Family socio-economic status remains a powerful predictor of achievement for both outcomes. I interpret my findings within three proposed mechanisms of parent engagement: cultivation ethic, realist reaction, and expressive logic.&#x0D; Keywords: parent engagement, literacy achievement, socio-economic status, inequality, summer learning, summer literacy camp(s)

Mix Contrast for COVID-19 Mild-to-Critical Prediction.

Objective: In a few patients with mild COVID-19, there is a possibility of the infection becoming severe or critical in the future. This work aims to identify high-risk patients who have a high probability of changing from mild to critical COVID-19 (only account for 5% of cases). Methods: Using traditional convolutional neural networks for classification may not be suitable to identify this 5% of high risk patients from an entire dataset due to the highly imbalanced label distribution. To address this problem, we propose a Mix Contrast model, which matches original features with mixed features for contrastive learning. Three modules are proposed for training the model: 1) a cumulative learning strategy for synthesizing the mixed feature; 2) a commutative feature combination module for learning the commutative law of feature concatenation; 3) a united pairwise loss assigning adaptive weights for sample pairs with different class anchors based on their current optimization status. Results: We collect a multi-center computed tomography dataset including 918 confirmed COVID-19 patients from four hospitals and evaluate the proposed method on both the COVID-19 mild-to-critical prediction and COVID-19 diagnosis tasks. For mild-to-critical prediction, the experimental results show a recall of 0.80 and a specificity of 0.815. For diagnosis, the model shows comparable results with deep neural networks using a large dataset. Our method demonstrates improvements when the amount of training data is small or imbalanced. Significance: Identifying mild-to-critical COVID-19 patients is important for early prevention and personalized treatment planning.