Guided Learning Strategy Research Articles

Rapid identification of Liubao tea vintage based on terahertz spectroscopy combined with improved differentiated creative search algorithm

As a special microbial fermented tea, the aging year of Liubao tea is a crucial determinant of its value. This study established a fast and high-precision method for identifying the age of Liubao tea by combining terahertz time-domain spectroscopy technology with chemometric methods. Most common optimization algorithms rely too much on the guidance of elite individuals in the optimization process and are prone to fall into local optimal solutions. Therefore, this paper uses a differentiated creative search algorithm with global thinking to optimize the support vector machine model parameters. To address the problem of poor search results due to unclear goals in the algorithm’s convergence and divergence processes, a guided learning strategy is employed to balance these schemes within the differentiated creative search algorithm. This approach yields a classification model with higher search efficiency. Compared with the support vector machine models optimized by Genetic Algorithm, Particle Swarm Optimization, and differentiated creative search algorithm, the new model achieved the best classification performance, with an accuracy of 96.87% and an F1 score of 0.9683. The results indicate that the guided learning strategy can balance the updating scheme of the differentiated creative search algorithm. The optimized model enables accurate qualitative analysis of the year of Liubao tea, offering a feasible solution for applying terahertz spectroscopy technology in tea identification.

Guided learning strategy: A novel update mechanism for metaheuristic algorithms design and improvement

Meta-heuristic algorithms (MH) are naturally inspired global optimization algorithms. They are often relatively simple and can solve problems in a short period of time, with certain benefits. However, as the problem becomes more complex, the solution that the algorithm can obtain is often not the optimal solution to the problem, which limits its application scenarios. Therefore, improving the performance and solving accuracy of existing algorithms is crucial for expanding their application ability. In traditional optimization algorithms, there are often two concepts, namely exploration and exploitation. Exploration refers to a wide range of discrete search, used to avoid falling into local optima, and exploitation refers to a small range of focused exploration, used to improve algorithm accuracy. How to balance exploration and exploitation is the key to enhancing algorithm performance and problem adaptability. This paper proposes a brand new strategy named Guided Learning Strategy (GLS) to solve above problem. GLS obtains the dispersion degree of the population by calculating the standard deviation of the historical locations of individuals in recent generations, and infers what guidance the algorithm currently needs. When the algorithm is biased towards exploration, it will guide the algorithm to exploit. Otherwise, it will guide the algorithm to explore. It is precisely because this strategy can identify the current needs of the algorithm and provide assistance that it can improve the performance of most algorithms. This article improves three types of algorithms based on evolution (LSHADE_SPACMA), Stochastic Fractal Search (SFS), and Marine Predators Algorithm (MPA) with better performance, and tests their performance on 57 constrained engineering problems and CEC2020. The effectiveness of this strategy has been confirmed and proved for optimization problem. The source codes of the proposed GLS (GLS_MPA) can be accessed by https://github.com/luchenghao2022/Guided-Learning-Strategy

Investment and Risk Management with Online News and Heterogeneous Networks

Stock price movements in financial markets are influenced by large volumes of news from diverse sources on the web, e.g., online news outlets, blogs, social media. Extracting useful information from online news for financial tasks, e.g., forecasting stock returns or risks, is, however, challenging due to the low signal-to-noise ratios of such online information. Assessing the relevance of each news article to the price movements of individual stocks is also difficult, even for human experts. In this article, we propose the Guided Global-Local Attention-based Multimodal Heterogeneous Network (GLAM) model, which comprises novel attention-based mechanisms for multimodal sequential and graph encoding, a guided learning strategy, and a multitask training objective. GLAM uses multimodal information, heterogeneous relationships between companies and leverages significant local responses of individual stock prices to online news to extract useful information from diverse global online news relevant to individual stocks for multiple forecasting tasks. Our extensive experiments with multiple datasets show that GLAM outperforms other state-of-the-art models on multiple forecasting tasks and investment and risk management application case-studies.

Hierarchical complementary residual attention learning for defocus blur detection

Defocus blur detection (DBD) aims to extract the in-focus part from a single image. Instead of adopting traditional hand-crafted features, recent deep neural networks based methods achieve DBD in an end-to-end architecture, obtaining many superior performances. However, these approaches are incapable of distinguishing many micro details when the image is complex. To tackle this issue, in this work, we propose Hierarchical Complementary Residual Attention Learning (HCRAL) for defocus blur detection, which employs multi-scale features with the complementary information from the network to guide the optimization of DBD. Specifically, to detect the micro details, residual attention is introduced to learn residual information between different scales, subsequently encouraging the network to re-focus on the easily-ignored boundary and sparse parts. Additionally, instead of only considering in-focus regions like other DBD methods, we capture the in-focus and out-of-focus features jointly. In this complementary mode, information ignored by one side may be learned by the other side, resulting in a positive influence between in-focus learning and out-of-focus learning. Simultaneously, the hierarchical feature guided learning strategy well leverages low- and high-level information and consequently enhances both global object detection and local boundary refinement. Experimental results on three datasets demonstrate the effectiveness and superiority of our methods.

A Novel Deep Learning Architecture with Multi-Scale Guided Learning for Image Splicing Localization

The goal of image splicing localization is to detect the tampered area in an input image. Deep learning models have shown good performance in such a task, but are generally unable to detect the boundaries of the tampered area well. In this paper, we propose a novel deep learning model for image splicing localization that not only considers local image features, but also extracts global information of images by using a multi-scale guided learning strategy. In addition, the model integrates spatial and channel self-attention mechanisms to focus on extracting important features instead of restraining unimportant or noisy features. The proposed model is trained on the CASIA v2.0 dataset, and its performance is tested on the CASIA v1.0, Columbia Uncompressed, and DSO-1 datasets. Experimental results show that, with the help of the multi-scale guided learning strategy and self-attention mechanisms, the proposed model can locate the tampered area more effectively than the state-of-the-art models.

Facilitating learning for students with special needs: a review of technology-supported special education studies

Owing to physical or mental disabilities, disabled students often come across more difficulties in learning. In an effort to improve their learning, researchers have adopted technology-supported tools to enhance disabled students’ adaptability to the learning environment and their learning achievement. The application of technology-supported special education has gradually increased in recent years. However, there is still a lack of investigation and analysis of the application and development trends of integrating technologies into special education. The aim of the present study was therefore to review technology-supported special education research articles by taking multiple dimensions into account, such as learning devices, learning strategies, learning domains and research issues, research subjects, types and level of disabilities, and learning environments. Based on the results, the number of studies has increased year by year, and the choice of learning devices and applications has become increasingly diverse; yet, the learning strategies still tend to be conservative since the majority of studies adopted the guided learning strategy. In addition, the application of technology has expanded to every learning domain, but is mainly focused on elementary school students and resource classrooms. Most importantly, the implementation of technology-supported special education tends not to result in teaching difficulties due to disabled students having different types and levels of disabilities. Further discussion and suggestions based on the findings can serve as a reference for teachers and researchers in special education.

Neural Bag-of-Ngrams

Bag-of-ngrams (BoN) models are commonly used for representing text. One of the main drawbacks of traditional BoN is the ignorance of n-gram's semantics. In this paper, we introduce the concept of Neural Bag-of-ngrams (Neural-BoN), which replaces sparse one-hot n-gram representation in traditional BoN with dense and rich-semantic n-gram representations. We first propose context guided n-gram representation by adding n-grams to word embeddings model. However, the context guided learning strategy of word embeddings is likely to miss some semantics for text-level tasks. Text guided n-gram representation and label guided n-gram representation are proposed to capture more semantics like topic or sentiment tendencies. Neural-BoN with the latter two n-gram representations achieve state-of-the-art results on 4 document-level classification datasets and 6 semantic relatedness categories. They are also on par with some sophisticated DNNs on 3 sentence-level classification datasets. Similar to traditional BoN, Neural-BoN is efficient, robust and easy to implement. We expect it to be a strong baseline and be used in more real-world applications.

Competencies acquisition through self‐directed learning among Malaysian managers

PurposeThe purpose of this paper is to examine how Malaysian managers acquire job competencies through self‐directed learning activities at their workplace. Specifically it aims to investigate what types of job competencies are required for the managers, how they learn to acquire those competencies, and whether the managers have the self‐directed learning attributes and capability to acquire job competencies by self‐directed learning activities.Design/methodology/approachThe data were collected through survey. The survey adapted the Self‐Directed Learning Readiness Scale (SDLRS) developed by Guglielmino. In total 238 respondents participated in this study.FindingsThe findings show that the respondents have identified communication, managerial, and job knowledge as the main competencies required for their jobs. Most of these competencies are acquired through on‐the‐job training, working as part of a team and self‐education. The findings also reveal that the respondents do possess the attributes of self‐directed learners.Research limitations/implicationsThe findings support previous research. Given the different cultural background of the Malaysian managers, their ways of learning are similar to those of Western managers. Future research can investigate how employees in other professions learn on their jobs and what other factors may influence self‐directed learning.Practical implicationsGiven the emphasis being placed on the importance of work place learning, it is timely to advance a workplace curriculum by integrating the everyday learning experiences and guided learning strategy in the organisation, so that the employees' learning can be facilitated more effectively.Originality/valueThe paper includes the different cultural context of how Malaysian managers learn.