A Survey of Deep Active Learning

Active learning (AL) attempts to maximize a model’s performance gain while annotating the fewest samples possible. Deep learning (DL) is greedy for data and requires a large amount of data supply to optimize a massive number of parameters if the model is to learn how to extract high-quality features. In recent years, due to the rapid development of internet technology, we have entered an era of information abundance characterized by massive amounts of available data. As a result, DL has attracted significant attention from researchers and has been rapidly developed. Compared with DL, however, researchers have a relatively low interest in AL. This is mainly because before the rise of DL, traditional machine learning requires relatively few labeled samples, meaning that early AL is rarely according the value it deserves. Although DL has made breakthroughs in various fields, most of this success is due to a large number of publicly available annotated datasets. However, the acquisition of a large number of high-quality annotated datasets consumes a lot of manpower, making it unfeasible in fields that require high levels of expertise (such as speech recognition, information extraction, medical images, etc.). Therefore, AL is gradually coming to receive the attention it is due. It is therefore natural to investigate whether AL can be used to reduce the cost of sample annotation while retaining the powerful learning capabilities of DL. As a result of such investigations, deep active learning (DeepAL) has emerged. Although research on this topic is quite abundant, there has not yet been a comprehensive survey of DeepAL-related works; accordingly, this article aims to fill this gap. We provide a formal classification method for the existing work, along with a comprehensive and systematic overview. In addition, we also analyze and summarize the development of DeepAL from an application perspective. Finally, we discuss the confusion and problems associated with DeepAL and provide some possible development directions.

Four principles for assessing student-directed projects

Student-directed projects are a promising approach to supporting powerful learning, yet uncertainty about how to assess these projects presents a barrier to widespread incorporation in K-12 classrooms. Drawing on interviews with computer science teachers and an interdisciplinary literature review, Karen Brennan, Sarah Blum-Smith, and Paulina Haduong offer four principles to guide assessment of student-directed projects: recognizing the individuality of the learner, illuminating process, engaging multiple perspectives, and cultivating capacity for personal judgment. They describe the research behind these principles and provide and example of what they look like in practice.

Creating something out of nothing: Enacting critical civic engagement in urban classrooms

Critical civic engagement (CCE) is a pedagogical framework for civic education in urban settings. CCE as a pedagogical approach engages student lived experience, develops critical thinking, and facilitates informed civic action projects. In this phenomenological study of teachers in four urban high schools in a large urban school district, the author seeks to understand how teachers experience the enactment of CCE elements in schools with majority African American or Latinx student populations. The author argues that CCE practices can and should lead to the development of civic identity as a critical outcome for students in contrast to more formal measures of academic achievement. Civic identity is the foundation upon with engagement in public life is built. The study suggests that the enactment of CCE elements provides a powerful learning and identity formation experience for students and a pedagogical process that inspires teachers and their student-centered practices.

Single session family therapy for beginners: what difference does it make to psychiatry registrars to participate in family sessions in front of and behind the screen?

Objective Current competencies required for fellowship of the RANZCP require psychiatry registrars to have experience in working with clients across all age groups, as well as working with families and the client’s wider network, however gaining this experience is not always easy for trainees. This paper reports on the experience of participating in Single Session Family Therapy (SSFT) during registrar training as a different modality for learning. Method: An online survey was conducted with fourteen registrars who had participated in SSFT during their child and adolescent rotation. Qualitative and simple quantitative data were collected and analysed. Results: Participating in SSFT during training was initially daunting, but had a positive effect on trainees, including influencing some towards focussing their future sub-specialisation in the child and youth area. Experience came through learning by doing, and seeing change. Registrars learnt about: understanding the role of the family; teamwork; technical skills; and gained confidence. Conclusions: Opportunities for trainees to participate in SSFT enables powerful learning beyond what can be taught in the classroom. Such opportunities may enhance registrars’ perceptions of family work, and may positively influence decision about future sub-specialisation.

Policy invariant explicit shaping: an efficient alternative to reward shaping

Reinforcement learning (RL) is a powerful learning paradigm in which agents can learn to maximize sparse and delayed reward signals. Although RL has had many impressive successes in complex domains, learning can take hours, days, or even years of training data. A major challenge of contemporary RL research is to discover how to learn with less data. Previous work has shown that domain information can be successfully used to shape the reward; by adding additional reward information, the agent can learn with much less data. Furthermore, if the reward is constructed from a potential function, the optimal policy is guaranteed to be unaltered. While such potential-based reward shaping (PBRS) holds promise, it is limited by the need for a well-defined potential function. Ideally, we would like to be able to take arbitrary advice from a human or other agent and improve performance without affecting the optimal policy. The recently introduced dynamic potential-based advice (DPBA) was proposed to tackle this challenge by predicting the potential function values as part of the learning process. However, this article demonstrates theoretically and empirically that, while DPBA can facilitate learning with good advice, it does in fact alter the optimal policy. We further show that when adding the correction term to “fix” DPBA it no longer shows effective shaping with good advice. We then present a simple method called policy invariant explicit shaping (PIES) and show theoretically and empirically that PIES can use arbitrary advice, speed-up learning, and leave the optimal policy unchanged.

Values-based Physical Education for the intermediate schooling phase in a diverse South African context

Participation in physical activity, Physical Education (PE) and sport has been recognised as a powerful learning tool for education, providing a universal language for contributing to valuable life principles. Values-based education implies that learners are educated about the aspects determining their behaviour. Values-based PE, physical activities and sport have the potential to transcend diversity and achieve cohesion, promote tolerance and trust and affirm respect between individuals and communities. The goal of PE can be to contribute to the acceptance of the infinite qualities of South Africa’s diversity and to claim the country’s diversity as a source of strength that forms a bond of a common set of values. There has been a global change in the interaction of learners with their environment; their lives are shaped by forces that do not necessarily assist them to learn and apply values. A PE programme infused with the values of Olympism and Ubuntuism can offer an investment in individual and societal improvement as the co-evolutionary interaction of these values and how they affect each learner can add to the celebration of human diversity. The question this study set out to answer was how can PE be used as a tool to teach values. Thus, the study aimed to inform the development of a values-based PE programme for the intermediate schooling phase. This qualitative study, from a constructivist paradigm, has enhanced the understanding of individuals’ cultures, beliefs and values, human experiences and situations. Purposeful sampling, of 10 intermediate phase teachers from five different public primary schools sought information-rich cases. The theoretical perspectives of the experiential learning theory were applied to teaching PE during in-service PE teacher training workshops. The process was documented by collecting data from multiple sources. Participatory action research was used, determining how data were collected, analysed and presented on an ongoing, cyclical basis. This study developed material for the intermediate phase PE curriculum that underpins the values of Olympism and Ubuntuism as core values, which were modelled by teachers and guided their work. The PE programme included key elements of and aligned with the study aims of the subject Life Skills. The outcomes of using PE as a tool to teach values propose recommendations to the Department of Basic Education of South Africa, to improve and implement a quality PE curriculum that is applicable to practice and that will optimise the chances of meeting National Curriculum Statement standards. Further research is recommended on the rest of the intermediate phase PE curriculum over the entire year, which includes other movement phenomena infused with values.

Giving feedback on peers’ concept maps as a learning experience: does quality of reviewed concept maps matter?

Giving feedback to peers can be a powerful learning tool because of the feedback provider’s active cognitive involvement with the products to be reviewed. The quality of peers’ products is naturally an important factor that might influence not only the quality of the feedback that is given, but also the learning arising from this process. This experimental study investigated the effect of the level of quality of the reviewed product on the knowledge acquisition of feedback providers, as well as the role of prior knowledge in this. Dutch secondary-school students (n = 77) were assigned to one of three conditions, which varied in the quality of the learning products (concept maps) on which students had to give feedback while working in an online physics inquiry learning environment. Post-test knowledge scores, the quality of students’ own concept maps and the quality of the feedback given were analyzed to determine any effect of condition on the learning of feedback providers. Students providing feedback on the lower-quality concept maps gave better feedback and had higher post-test scores. There was no interaction with level of prior knowledge. Possible implications for practice and further research directions are discussed.

Recommender System for Telecom Packages Based on the Deep & Cross Network

With the evolution of 5G technology, the telecom industry has influenced the livelihood of people and impacted the development of the national economy significantly. To increase revenue per customer and secure long-term contracts of users, telecommunications firms and enterprises have launched different types of telecom packages to satisfy the varying requirements of users. Although several recommender systems have been proposed in recent years for telecommunication package recommendation, extracting effective feature information from large and complex consumption data remains challenging. Considering the telecom package recommendation problems, traditional recommendation methods either use complex expert feature engineering or fail to perform end-to-end deep learning training. In this study, a recommender system based on deep & cross network (DCN), deep belief network (DBN), Embedding, and Word2Vec is proposed using the powerful learning abilities of deep learning. The proposed system fits the telecom package recommender system in terms of click-through rate prediction to provide a potential solution for the recommendation challenges faced by telecom enterprises. The proposed model can effectively capture the finite order interactional features and deep hidden features. Additionally, the text information in the data is completely used to further improve the recommendation ability of the model. Moreover, the proposed method does not require feature engineering. We conducted comprehensive experiments using real-world datasets, and the results verify that our method can generate improved recommendation accuracy in comparison with those observed in DBN, DCN, deep factorization machine, and deep neural network models individually.

Teaching palliative care skills via simulation-based learning

Background: The scarcity of palliative care (PC) services in Morocco, and their absence in Settat, limits the opportunities for nursing students at the Higher Institute of Health Sciences (HIHS) to benefit from clinical placements. As a consequence of this, most students feel underprepared to care for patients with PC needs. Aim: The purpose of this study is to share a simulation-based learning experience in a PC context and to evaluate the effectiveness of this learning method. Methods: The simulation experience took place in the simulation centre of the HIHS and involved 20 nursing students in their second year. The main goal of the simulation session was to simulate the support given to patients going through the five stages of grief. A post-simulation survey was conducted to explore the nursing students reflections on this learning experience. Findings: The simulation is recommended as a powerful learning approach to compensate for the lack of PC clinical placements available to nursing students. Conclusion: The simulation-based training was an excellent opportunity for nursing students to experience caring for patients in extreme end-of life-situations, which was not possible before due to the lack of specialised PC services.

How to Improve the Actual Effect of Computer-Assisted Teaching

Clark & Starr (1986) found in a study that the amount of memory of students varies depending on the situation: 10% of people can remember the information or knowledge “read”; 20% can remember the “heard” Information; 30% of people can remember what they “see”; 50% of people can remember what they “hear and see”; 70% of people can remember information “said”; 90% can Remember the “said and done” things. Therefore, if the learner only uses reading or listening or seeing methods when studying, the amount of memory of information is limited, but if the learner can do it by hand, it can deepen the impression and increase the memory capacity of information. Although traditional textbooks can assist learners to record external information in words, they still have certain limitations on abstract mathematical knowledge or scientific learning, and the rapid development of computers can make up for this deficiency. Computers can present dynamic images and provide learners with a powerful learning and perceptual experience, enabling learners to be more perceptive to abstract concepts (Fan, 2014).