Learning In Complex Research Articles

Weed detection using deep learning in complex and highly occluded potato field environment

Weed management is a significant challenge for agronomists, especially in highly dense field environments. The study aims to develop a computer vision-based system for distinguishing between potato plants and weeds in the post-emergence stage within high occlusion and complex backgrounds. To achieve this, the RGB dataset having real images was collected from potato farms. After thorough cleaning, the data augmentation was done. Images are annotated at the pixel level and are made freely available to the research community for future investigations. To accurately segment, detect, and classify the potato plant from the weeds, cutting-edge deep learning techniques, i.e., the Mask RCNN and YOLO version 8 (YOLOv8), are trained. Precision (P), recall (R), and mean average precision (mAP@0.5, mAP@0.50-0.95) were used to evaluate the different model scales. YOLOv8 obtained the mAP@0.50 83.4%, whereas Mask RCNN stood at 79%. Mask RCNN obtained the highest P(0.83), R(0.76), and F1(0.91) values for the weed class. The reported performance metrics indicate that although YOLOv8 slightly outperforms Mask RCNN in overall mAP, Mask RCNN achieves higher precision, recall, and F1 scores for the weed class. This implies that Mask RCNN may be more effective at identifying weeds, which is crucial for effective weed management. These accuracy levels may not be exceptional, but they are a testament to the model's capabilities for weed detection in a highly complex and heavily occluded environment.

An Effective Federated Object Detection Framework with Dynamic Differential Privacy

The proliferation of data across multiple domains necessitates the adoption of machine learning models that respect user privacy and data security, particularly in sensitive scenarios like surveillance and medical imaging. Federated learning (FL) offers a promising solution by decentralizing the learning process, allowing multiple participants to collaboratively train a model without sharing their data. However, when applied to complex tasks such as object detection, standard FL frameworks can fall short in balancing the dual demands of high accuracy and stringent privacy. This paper introduces a sophisticated federated object detection framework that incorporates advanced differential privacy mechanisms to enhance privacy protection. Our framework is designed to work effectively across heterogeneous and potentially large-scale datasets, characteristic of real-world environments. It integrates a novel adaptive differential privacy model that strategically adjusts the noise scale during the training process based on the sensitivity of the features being learned and the progression of the model’s accuracy. We present a detailed methodology that includes a privacy budget management system, which optimally allocates and tracks privacy expenditure throughout training cycles. Additionally, our approach employs a hybrid model aggregation technique that not only ensures robust privacy guarantees but also mitigates the degradation of object detection performance typically associated with DP. The effectiveness of our framework is demonstrated through extensive experiments on multiple benchmark datasets, including COCO and PASCAL VOC. Our results show that our framework not only adheres to strict DP standards but also achieves near-state-of-the-art object detection performance, underscoring its practical applicability. For example, in some settings, our method can lower the privacy success rate by 40% while maintaining high model accuracy. This study makes significant strides in advancing the field of privacy-preserving machine learning, especially in applications where user privacy cannot be compromised. The proposed framework sets a new benchmark for implementing federated learning in complex, privacy-sensitive tasks and opens avenues for future research in secure, decentralized machine learning technologies.

Youth and climate crisis: participation and learning in complex, adaptive systems

AbstractThe Friday for Future school strikes have vividly demonstrated young people’s concern for the rapidly deteriorating life and survival chances of future generations of human and nonhuman beings. The young people’s ambivalence between despair and hope, feelings of powerlessness and empowerment challenge us to recognise the complexity and dynamic not only of ecosystems but of socio-ecological systems and to question the mechanistic and linear thinking often prevalent in youth research and education science. This thinkbite invites readers to consider young people’s experiences with the climate crisis and (climate change) education through the perspective of Complex Adaptive System Theory to open up learning space for more wholesome and creative responses to the challenge.

Federated BrainAge estimation from MRI: a proof of concept

AbstractBackgroundThe gap between chronological and biological brain‐age, represented by MRI features, associates with dementia risk (Wang et. al., He et. al.). Age‐scores are predicted, however, using deep learning models, which require large amounts of MRI data. This poses a challenge, as patient brain MRI data is highly privacy‐sensitive and legal barriers often impede the sharing or transferring of data. Federated learning provides a potential solution by enabling training a global model on distributed data, allowing biobanks to keep data local while creating a shared model. In this study, we assess whether a federated set‐up is feasible for brain‐age estimation.MethodWe set up a federated learning framework across three biobanks (Fig. 2) to train a BrainAge model (Wang et. al). Data of healthy individuals from the ADNI dataset was randomly split across the three biobanks (N = 221, 219, 217). MRI images were pre‐processed into gray matter density maps (Good et. al.). The model was iteratively trained using federated averaging (McMahan et. al.) on data at two biobanks (70/30 training/validation split), and tested at the remaining biobank. We trained the model for ten rounds, and optimized for the number of epochs, learning rate, and learning decay hyperparameters. These results were compared to a centrally trained model, where the training samples were pooled together. Additionally, we applied the trained model to individuals with Mild Cognitive Impairment and Alzheimer’s Disease and performed a survival analysis. We assessed performance using Mean Absolute Error (MAE).ResultFederated learning (MAE = 4.02) achieves similar test performance as central learning (MAE = 3.94) (Table 1). The model yielded an MAE of 4.89 for MCI and 5.45 for AD patients; this increase may be explained by their neurodegeneration. We also see a statistically different survival curve for developing AD in individuals with a brain‐age‐gap (Fig. 1), displaying an ‘older’ profile for their chronological age. This might indicate higher risk of developing AD, in line with previous results (Wang et. al.).ConclusionResults show federated learning has the potential to address the challenge of data accessibility for machine learning in complex and data hungry research, including dementia.

Simultaneous fMRI and tDCS for Enhancing Training of Flight Tasks.

There is a gap in our understanding of how best to apply transcranial direct-current stimulation (tDCS) to enhance learning in complex, realistic, and multifocus tasks such as aviation. Our goal is to assess the effects of tDCS and feedback training on task performance, brain activity, and connectivity using functional magnetic resonance imaging (fMRI). Experienced glider pilots were recruited to perform a one-day, three-run flight-simulator task involving varying difficulty conditions and a secondary auditory task, mimicking real flight requirements. The stimulation group (versus sham) received 1.5 mA high-definition HD-tDCS to the right dorsolateral prefrontal cortex (DLPFC) for 30 min during the training. Whole-brain fMRI was collected before, during, and after stimulation. Active stimulation improved piloting performance both during and post-training, particularly in novice pilots. The fMRI revealed a number of tDCS-induced effects on brain activation, including an increase in the left cerebellum and bilateral basal ganglia for the most difficult conditions, an increase in DLPFC activation and connectivity to the cerebellum during stimulation, and an inhibition in the secondary task-related auditory cortex and Broca's area. Here, we show that stimulation increases activity and connectivity in flight-related brain areas, particularly in novices, and increases the brain's ability to focus on flying and ignore distractors. These findings can guide applied neurostimulation in real pilot training to enhance skill acquisition and can be applied widely in other complex perceptual-motor real-world tasks.

Communication-Efficient Distributed Learning Over Networks—Part II: Necessary Conditions for Accuracy

Distributed learning is crucial for many applications such as localization and tracking, autonomy, and crowd sensing. This paper investigates communication-efficient distributed learning of time-varying states over networks. Specifically, the paper considers a network of nodes that infer their current states in a decentralized manner using observations obtained via local sensing and messages obtained via noisy inter-node communications. The paper derives a necessary condition in terms of the sensing and communication capabilities of the network for the boundedness of the learning error over time. The necessary condition is compared with the sufficient condition established in a companion paper and the gap between the two conditions is discussed. The paper provides guidelines for efficient management of the sensing and communication resources for distributed learning in complex networked systems.

Communication-Efficient Distributed Learning Over Networks—Part I: Sufficient Conditions for Accuracy

Distributed learning is an important task in emerging applications such as localization and navigation, Internet-of-Things, and autonomous vehicles. This paper establishes a theoretical framework for learning states that evolve in real time over networks. Specifically, each agent node in the network aims to infer a time-varying state in a decentralized manner by using the node’s local observations and the messages received from other nodes within its communication range. As a result, the inference accuracy of a node is significantly affected by the quality of its received messages. This calls for carefully designed strategies for generating messages that are able to provide sufficient information for the receiver and are robust to channel impairments. This paper presents communication-efficient encoding strategies for generating transmitted messages and derives a sufficient condition for the boundedness of the distributed inference error of all the agent nodes over time. The findings of this paper provide guidelines for the design of communication-efficient distributed learning in complex networked systems.

Mapping Health Professions Education: Using Complexity Science to Make Sense of Learning Through Electronic Consultations.

Phenomena studied within health professions education are often complex and multifaceted. This article describes a complexity science-informed theoretical framework that was developed for exploring how electronic consultations support learning among primary care providers, as well as within the larger organizations or systems in which they practice. This framework enables researchers to investigate learning occurring simultaneously at multiple levels (including individuals and social groups), without simplistically conflating levels or theories. The various levels of learning and associated theories are illustrated using examples from electronic consultations. This complexity science-inspired framework can be used for studying learning in complex, multilayered systems.

Learning multi-agent cooperation.

Advances in reinforcement learning (RL) have resulted in recent breakthroughs in the application of artificial intelligence (AI) across many different domains. An emerging landscape of development environments is making powerful RL techniques more accessible for a growing community of researchers. However, most existing frameworks do not directly address the problem of learning in complex operating environments, such as dense urban settings or defense-related scenarios, that incorporate distributed, heterogeneous teams of agents. To help enable AI research for this important class of applications, we introduce the AI Arena: a scalable framework with flexible abstractions for associating agents with policies and policies with learning algorithms. Our results highlight the strengths of our approach, illustrate the importance of curriculum design, and measure the impact of multi-agent learning paradigms on the emergence of cooperation.

Dynamics of self­esteem quality of life by health workers during the Coronavirus pandemic in 2020—2021

A relatively new term, VUCA, has emerged in the modern information field. It clearly reflects 4 essential features of modern life: Volatility — due to the need to study many factors that change in the decision-making process; Uncertainty — the inability to see and predict the future prospects; Complexity — lack of clarity of understanding of many events; Ambiguity — uncertainty of interpretation of the situation. This could not but affect the self-esteem of young health professionals of their quality of life (QOL) during 2020 and 2021.
 Objective — to remotely monitor the dynamics of changes in self-esteem of certain indicators of quality of life (QOL) in young health professionals depending on their professional background against the negative impact of chronic stress (infodemia), various administrative quarantine restrictions and forced professional communication with various patients with somatic diseases. during the COVID-19 pandemic during 2020—2021.
 Materials and methods. The study used: general clinical methods — for the overall assessment of somatic condition; psychodiagnostic survey – adapted SF-36 questionnaire to determine QOL; biostatistical methods — for the purpose of mathematical processing of the received data.
 Results and discussion. Interpersonal contacts with 4th year students of medical and dental faculties related to remote testing for the period from spring 2020 to December 2021 took place exclusively remotely due to quarantine restrictions and mainly at the initiative of health workers themselves, who learned about the possibility to effectively assess their psychological state and QOL.An objective analysis of the dynamics of self-assessment by young health workers of their own QOL over the past two years shows their successful psychosomatic adaptation to learning and clinical work in complex, unpredictable due to instability, ambiguity and some uncertainty in the 2020—2021 pandemic. Their average self-assessment of QOL should be considered as a kind of positive standard control when compared with data from similar tests conducted among other segments of the population.
 Conclusions. An objective analysis of the dynamics of self-esteem by young physicians of their own QOL over the past two years shows their successful psychosomatic adaptation in the pandemic of 2020—2021. They should be considered as a kind of positive standard when compared with data from similar testing of other segments of the population in these years.

A model of mood as integrated advantage.

Mood is an integrative and diffuse affective state that is thought to exert a pervasive effect on cognition and behavior. At the same time, mood itself is thought to fluctuate slowly as a product of feedback from interactions with the environment. Here we present a new computational theory of the valence of mood-the Integrated Advantage model-that seeks to account for this bidirectional interaction. Adopting theoretical formalisms from reinforcement learning, we propose to conceptualize the valence of mood as a leaky integral of an agent's appraisals of the Advantage of its actions. This model generalizes and extends previous models of mood wherein affective valence was conceptualized as a moving average of reward prediction errors. We give a full theoretical derivation of the Integrated Advantage model and provide a functional explanation of how an integrated-Advantage variable could be deployed adaptively by a biological agent to accelerate learning in complex and/or stochastic environments. Specifically, drawing on stochastic optimization theory, we propose that an agent can utilize our hypothesized form of mood to approximate a momentum-based update to its behavioral policy, thereby facilitating rapid learning of optimal actions. We then show how this model of mood provides a principled and parsimonious explanation for a number of contextual effects on mood from the affective science literature, including expectation- and surprise-related effects, counterfactual effects from information about foregone alternatives, action-typicality effects, and action/inaction asymmetry. (PsycInfo Database Record (c) 2022 APA, all rights reserved).

Scrum methodology in context-based secondary chemistry classes: effects on students\u2019 achievement and on students\u2019 perceptions of affective and metacognitive dimensions of their learning

Teaching with Scrum methodology includes ceremonies, roles and artefacts supporting students in planning, monitoring and directing their learning process. It scaffolds students’ learning in complex and sometimes overwhelming context-based learning environments. Effects of the implementation on both students’ learning outcomes and self-reported perceptions of six affective and metacognitive outcomes were investigated. Six teachers implemented Scrum methodology in a context-based secondary chemistry module on Green Chemistry. Their classes formed the experimental group. Based on how students experienced the quality of the implementation, teachers of the experimental group were subdivided into top-teachers and growth-teachers. Consequently, their students formed two sub-experimental groups. The comparison group, which did not use Scrum methodology, consisted of students taught by four teachers. A pre-test post-test control group design was used to study its effect on students’ achievements and self-reported affective and metacognitive outcomes. Students of both experimental groups outperformed students of the comparison group with a large effect-size (top-teachers); and medium effect-size (growth-teachers) on learning outcomes. Findings on students’ perceptions of affective and metacognitive outcomes revealed medium and small effects of Scrum methodology. Despite the fact that the implementation is challenging for teachers, it appears that Scrum methodology has positive effects on students’ achievement and on students’ perceptions of affective and metacognitive dimensions of their learning.

Balancing Collaboration and Cognitive Load to Optimize Individual and Group Desirable Difficulties

We sought to identify factors that optimize individual learning in complex, technology-enhanced learning environments. Undergraduates viewed tutorials and played a simulation-based game either alone or in groups and in either high or low cognitive load sequences and later took tests measuring comprehension of tutorials and transfer of computer networking skills. A cognitive load by collaboration interaction was found for both immediate and delayed transfer measures, but not comprehension measures. Students working in groups performed best under high cognitive load whereas students working individually performed best under low cognitive load. These findings support the notions of optimal individual and group cognitive load and have implications for leveraging technology to design learning environments that allow students to collaborate and maximize individual learning.

AFTER THE DELUGE: AN ACTION NOTEBOOK FOR A RESPONSIBLE SOCIOLINGUIST

In this paper, we begin with inevitability and cyclicity of pandemic during the last two centuries. The problems seem to be coming back to us again and again through Small-pox, Tuberculosis, Plague, Influenza, HIV/AIDS, Ebola, and now COVID-19. The 2003 Severe Acute Respiratory Syndrome (SARS) pandemic and the Avian Influenza resulted in establishment of WHO protocols and awareness. However, as we consider the people at the Bottom of the Pyramid in developing and poor countries, the problem of survival of ethnic groups and their languages appear to be more acute. We know that 96% of India’s mother-tongues are spoken by only 4% of the population. But as India houses 3592 numerically weak mother tongues (spoken by 705 tribes or ethnic groups and 1284 castes scattered all over our rural landscape), with high degree of poverty combined with social hierarchies based on caste, religion and ethnicity, the shape of our population pyramid is relatively flat and bulging at the bottom making the problem of the survival of mother-tongues and learning in other-tongues complex. Our problem is that these ethnic groups and speakers of indigenous languages are fast losing their grip over their respective mother-tongues as they have to survive in an urban cauldron. It is this group that faced/faces the COVID-19 crisis more than anyone else, and it is this group that was seen walking back from urban centres to their villages. But those who stayed back in villages are also vulnerable to the pandemic. The effect of COVID-19 on our indigenous groups could be dangerous. After defining and describing ‘Danger’, we argue that the creativity of these indigenous language speakers is facing a grave threat. This is not only a matter of survival or production, but also an issue of printing, publishing, being read, translated, and sold. After discussing a few theoretical positions, from Ethnologue to Fishman and others, a list of tasks of Responsible Sociolinguistics is enlisted here. Then a detailed comparison is held between Biological and Linguistic Endangerment. What lessons could be learned by sociolinguists and language planners is discussed. The kind of tensions existing as real threat to survival are discussed in terms of systems theory and power-conflict theory. We argue that Linguistics should be used for healing of the wounds and injured pride of the smaller speech groups.

Evaluation of Instance-Based Learning and Q-Learning Algorithms in Dynamic Environments

Reinforcement learning is an unsupervised learning algorithm, where learning is based upon feedback from the environment. Prior research has proposed cognitive (e.g., Instance-based Learning or IBL) and statistical (Q-learning) reinforcement learning algorithms. However, an evaluation of these algorithms in a single dynamic environment has not been explored. In this paper, a comparison between the statistical Q-learning algorithm and the cognitive IBL algorithm is presented. A well-known environment, “Frozen Lake,” is used to train, generalize, and scale Q-learning and IBL algorithms. For generalizing, the Q-learning and IBL agents were trained on one version of the Frozen Lake and tested on a permuted version of the same environment. For scaling, the two algorithms were tested on a larger version of the Frozen Lake environment. Results revealed that the IBL algorithm used less training time and generalized better to different environment variants. The IBL algorithm was also able to show scalability by retaining its superior performance in the larger environment. These results indicate that the IBL algorithm could be proposed as an alternative to the standard reinforcement learning algorithms based on dynamic programming such as Q-learning. The inclusion of human factors (such as memory) in the IBL algorithm makes it suitable for robust learning in complex and dynamic environments.

Horseshoe Regularisation for Machine Learning in Complex and Deep Models1

SummarySince the advent of the horseshoe priors for regularisation, global–local shrinkage methods have proved to be a fertile ground for the development of Bayesian methodology in machine learning, specifically for high‐dimensional regression and classification problems. They have achieved remarkable success in computation and enjoy strong theoretical support. Most of the existing literature has focused on the linear Gaussian case; for which systematic surveys are available. The purpose of the current article is to demonstrate that the horseshoe regularisation is useful far more broadly, by reviewing both methodological and computational developments in complex models that are more relevant to machine learning applications. Specifically, we focus on methodological challenges in horseshoe regularisation in non‐linear and non‐Gaussian models, multivariate models and deep neural networks. We also outline the recent computational developments in horseshoe shrinkage for complex models along with a list of available software implementations that allows one to venture out beyond the comfort zone of the canonical linear regression problems.

Learning in complex, multi-component cognitive systems: Different learning challenges within the same system.

Using word learning as an example of a complex system, we investigated how differences in the structure of the subcomponents in which learning occurs can have significant consequences for the challenge of integrating new information within such systems. Learning a new word involves integrating information into the two key stages/subcomponents of processing within the word production system. In the first stage, multiple semantic features are mapped onto a single word. Conversely, in the second stage, a single word is mapped onto multiple segmental features. We tested whether the unitary goal of word learning leads to different local outcomes in these two stages because of their reversed mapping patterns. Neurotypical individuals (N = 17) learned names and semantic features for pictures of unfamiliar objects presented in semantically related, segmentally related and unrelated blocks. Both similarity types interfered with word learning. However, feature learning was differentially affected within the two subcomponents of word production. Semantic similarity facilitated learning distinctive semantic features (i.e., features unique to each item), whereas segmental similarity facilitated learning shared segmental features (i.e., features common to several items in a block). These results are compatible with an incremental learning model in which learning not only strengthens certain associations but also weakens others according to the local goals of each subcomponent. More generally, they demonstrate that the same overall learning goal can lead to opposite learning outcomes in the subcomponents of a complex system. The general principles uncovered may extend beyond word learning to other complex systems with multiple subcomponents. (PsycINFO Database Record (c) 2019 APA, all rights reserved).

Supplemental Material for Learning in Complex, Multi-Component Cognitive Systems: Different Learning Challenges Within the Same System

Supplemental Material for Learning in Complex, Multi-Component Cognitive Systems: Different Learning Challenges Within the Same System

Action Learning Questions: Making Sense of Organizational Chaos

This article uses a dialogic and questioning approach between the two authors, to explore how action learning questioning as a process has been used to create transformation and capability in some organizations impacted by massive change. Action Learning, based on the original work of Reg Revans, is a question driven approach enabling complex business issues to be explored through action and reflection with fellow set members. The authors narrate how such an approach has been used as the foundation for programs of professional development qualifications, with the assistance of some universities willing to forego their usual practice of teaching and testing concepts and theories, acknowledging that these new business development programs can succeed in bringing about results, resilience and competence in workforces battered by change fatigue and chaos. Too often organizational development concepts have become dry theoretical concepts in text books. This exchange demonstrates how they can come to life through questioning, reflective action and organizational support using evidence from major programs now underway in government and commercial sectors in the UK. The authors demonstrate how Action Learning Questioning based initiatives have the capacity to build social capital, and generate effective learning in complex and rapidly changing business and organizational environments. It concludes with suggestions for the future.

Using mobile technology for workplace learning

Students' agency is an importnat enabler of productive learning in complex, unpredictable workplace environments. In the study presented here, we explored how mobile technology can help students enhance their workplace learning experiences and develop their capacity to act as learners and future practitioners. We collected survey and interview data from 312 participants, which informed the development of Mobile Technology Capacity Building Framework that comprises thematic resources for students, academics and workplace educators. Its development draws on two sets of theoretical ideas: the importance of agentic learning that enables students to develop their practice capabilities; and the use of activity-centred learning design to distinguish between what can be designed ahead of time and what should be left to students' agency. This study and Framework contribute to understanding how the productive use of technologies can foster students' agency and development of deliberate professionals with a high sense of adaptive expertise.