Adaptive Selection Research Articles

Optimal layer selection for latent data augmentation

While data augmentation (DA) is generally applied to input data, several studies have reported that applying DA to hidden layers in neural networks, i.e., feature augmentation, can improve performance. However, in previous studies, the layers to which DA is applied have not been carefully considered, often being applied randomly and uniformly or only to a specific layer, leaving room for arbitrariness. Thus, in this study, we investigated the trends of suitable layers for applying DA in various experimental configurations, e.g., training from scratch, transfer learning, various dataset settings, and different models. In addition, to adjust the suitable layers for DA automatically, we propose the adaptive layer selection (AdaLASE) method, which updates the ratio to perform DA for each layer based on the gradient descent method during training. The experimental results obtained on several image classification datasets indicate that the proposed AdaLASE method altered the ratio as expected and achieved high overall test accuracy.

A denoising algorithm based on ARIMA and competitive K-SVD for the diagnosis of rolling bearing faults

Rolling bearings are extensively employed in industrial production as essential support components for rotating machinery. However, under conditions of high load and harsh operation, bearings are highly susceptible to failure. The weak vibration signals associated with these failures may be obscured by complex harmonic interference and strong noise, posing challenges for the accurate diagnosis of rolling bearing failures. In this paper, an autoregressive integrated moving average and competitive K-singular value decomposition (ARIMA-CK-SVD) algorithm is proposed to realize effective extraction of faulty pulse signals in a strong interference environment. First, the ARIMA model is used to preprocess the original signal to eliminate the interference of harmonic components. Second, a method is proposed for the adaptive selection of parameters in the ARIMA model, with consideration given to the characteristics of K-SVD. Subsequently, a competitive mechanism is introduced during the dictionary update phase of the algorithm to adjust the pattern of atomic updates and eliminate noise atoms. Finally, the effectiveness of the ARIMA-CK-SVD has been validated through simulation experiments and engineering tests.

Microhabitat influences on phage-bacteria dynamics in an abandoned mine for ecorestoration

Understanding the complex interactions between bacteriophages (phages) and bacteria within varied environmental niches is critical yet underexplored for improving microbe-assisted ecological restoration. This study investigates the influence of microhabitat heterogeneity within an abandoned mine on phage-bacteria interaction patterns, focusing on Pseudomonas-enriched bacterial communities. By isolating viral communities and purifying bacteria from soils of three distinct microhabitats, we assessed the regulatory role of environmental factors on these interactions, crucial for bacterial success in environmental applications. We characterized microhabitat variability by analyzing soil particle size fractions, minerals composition, and elemental content using X-ray diffraction and energy-dispersive X-ray analyses. 16S rRNA sequencing and cross-infection assays revealed that although bacterial communities across different microhabitats are taxonomically similar, their interaction patterns with phages are distinct. Phage communities showed nonselective infectivity across soil types, while bacterial communities exhibited selective adaptation, facilitating colonization across diverse microhabitats. Minerals such as mica, kaolinite, and hematite were found to increase phage infectivity, whereas mixed-layer clay correlated with early lysis. Additionally, higher levels of iron (Fe) and potassium (K) were linked to bacterial resistance strategies. Our findings highlight the importance of understanding asymmetric adaptive strategies between bacteria and phages, driven by microhabitat heterogeneity, for enhancing microbial-mediated nature-based restoration of degraded ecosystems.

Multi-source Selective Graph Domain Adaptation Network for cross-subject EEG emotion recognition

Affective brain-computer interface is an important part of realizing emotional human–computer interaction. However, existing objective individual differences among subjects significantly hinder the application of electroencephalography (EEG) emotion recognition. Existing methods still lack the complete extraction of subject-invariant representations for EEG and the ability to fuse valuable information from multiple subjects to facilitate the emotion recognition of the target subject. To address the above challenges, we propose a Multi-source Selective Graph Domain Adaptation Network (MSGDAN), which can better utilize data from different source subjects and perform more robust emotion recognition on the target subject. The proposed network extracts and selects the individual information specific to each subject, where public information refers to subject-invariant components from multi-source subjects. Moreover, the graph domain adaptation network captures both functional connectivity and regional states of the brain via a dynamic graph network and then integrates graph domain adaptation to ensure the invariance of both functional connectivity and regional states. To evaluate our method, we conduct cross-subject emotion recognition experiments on the SEED, SEED-IV, and DEAP datasets. The results demonstrate that the MSGDAN has superior classification performance.

Self-regulated learning processes of advanced musicians: A PRISMA review

This systematic review aimed to provide a comprehensive and up-to-date understanding of the current state of empirical research in self-regulated learning (SRL) and music, investigating learning strategies planned and employed by advanced musicians and evidence to support the benefits of SRL training for this population. Using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) protocol, we initially identified 3323 records related to our search, spanning the period from 2001 to 2023. Inclusion criteria consisted of peer-reviewed empirical articles reporting studies involving the participation of advanced (i.e., undergraduate and postgraduate student and professional) musicians using Zimmerman’s cyclical model of SRL and published in English, Spanish, Portuguese, or French. Forty-eight articles met all the inclusion criteria and were included in the review. We divided them into three types: reports of observational studies, descriptive-correlational, and intervention studies. We discuss the findings according to the three phases of Zimmerman’s cyclical model: forethought, performance/volitional control, and self-reflection, including the processes of goal setting, selection and adaptation of learning strategies, help seeking, time management, and self-evaluation and adaptation. Our analysis identified trends in the behavior of advanced musicians, particularly regarding practice quantity and musical achievement. Suggestions for future research indicate the need to invest in the training of instrumental and vocal teachers, so that they become familiar with SRL learning strategies that they can pass on to their students.

ESCAPE: Energy-based Selective Adaptive Correction for Out-of-distribution 3D Human Pose Estimation

Despite recent advances in human pose estimation (HPE), poor generalization to out-of-distribution (OOD) data remains a difficult problem. While previous works have proposed Test-Time Adaptation (TTA) to bridge the train-test domain gap by refining network parameters at inference, the absence of ground-truth annotations makes it highly challenging and existing methods typically increase inference times by one or more orders of magnitude. We observe that (1) not every test time sample is OOD, and (2) HPE errors are significantly larger on distal keypoints (wrist, ankle). To this end, we propose ESCAPE: a lightweight correction and selective adaptation framework which applies a fast, forward-pass correction on most data while reserving costly TTA for OOD data. The free energy function is introduced to separate OOD samples from incoming data and a correction network is trained to estimate the errors of pretrained backbone HPE predictions on the distal keypoints. For OOD samples, we propose a novel self-consistency adaptation loss to update the correction network by leveraging the constraining relationship between distal keypoints and proximal keypoints (shoulders, hips), via a second “reverse” network. ESCAPE improves the distal MPJPE of five popular HPE models by up to 7% on unseen data, achieves state-of-the-art results on two popular HPE benchmarks, and is significantly faster than existing adaptation methods.

Rihi Mesh Placement Algorithm In Collocation Methods

Various adaptive mesh selection strategies for solving single higher order two-points boundary value problems (BVPs) by using collocation methods are intensively investigated for along time and they are now well established. In this work we concern with numerical investigations of adaptive mesh selection algorithms using the criterion function rihi for solving first order system of BVPs and developing some algorithms. The algorithms perform quite nicely and appear competitive with De Boor algorithm.

Selection on the vascular-remodeling BMPER gene is associated with altitudinal adaptation in an insular lizard

Abstract High altitude imposes several extreme constraints on life, such as low oxygen pressure and high levels of ultraviolet radiation, which require specialized adaptations. Many studies have focused on how endothermic vertebrates respond to these challenging environments, but there is still uncertainty on how ectotherms adapt to these conditions. Here, we used whole-genome sequencing of low-altitude (100–600 m) and high-altitude (3,550 m) populations of the wide-ranging Tenerife lizard Gallotia galloti to uncover signatures of selection for altitudinal adaptation. The studied populations show reduced differentiation, sharing similar patterns of genetic variation. Selective sweep mapping suggests that signatures of adaptation to high altitude are not widespread across the genome, clustering in a relatively small number of genomic regions. One of these regions contains BMPER, a gene involved with vascular remodeling, and that has been associated with hypoxia-induced angiogenic response. By genotyping samples across 2 altitudinal transects, we show that allele frequency changes at this locus are not gradual, but rather show a well-defined shift above ca. 1,900 m. Transcript and protein structure analyses on this gene suggest that putative selection likely acts on noncoding variation. These results underline how low oxygen pressure generates the most consistent selective constraint in high-altitude environments, to which vertebrates with vastly contrasting physiological profiles need to adapt in the context of ongoing climate change.

An evolutionary perspective on capabilities for fluid product-markets: The contingent effects of routinization and renewal in marketing, R&D, and operations

The performance benefits of functional capabilities in marketing, technology, and operations rely on their routinization in organizational processes, but these also require renewal in response to environmental change. This raises a fundamental tension: is it better to maximally develop functional capabilities that offer the highest contingent benefit in present market conditions, and/or to modify capabilities as conditions change? We propose two measures of a firm's ability to renew its functional capabilities to align with market conditions: capability heterogeneity (variation in extant capabilities) and capability adaptability (selection among these strategic options). In a 20-year panel of 771 firms, we find environmental change increases the importance of these aspects of how capabilities are managed relative to what capabilities a firm possesses: In stable product-markets, capability heterogeneity and adaptability incur significant costs whereas functional capabilities improve profitability. In contrast, functional capabilities can be detrimental in fluid product-markets whereas heterogeneity and adaptability increase profitability. Notably, marketing capability remains beneficial across environments, acting as a profitable alternative to capability heterogeneity and adaptability when future conditions are uncertain. This evolutionary perspective contributes to ongoing theoretical debates on the conceptualization and consequences of capabilities, with practical implications for mitigating the risks of excessive inertia or change.

Normalizing flow regularization for photoacoustic tomography

Abstract Proper regularization is crucial in inverse problems to achieve high-quality reconstruction, even with an ill-conditioned measurement system. This is particularly true for three-dimensional photoacoustic tomography, which is computationally demanding and requires rapid scanning, often leading to incomplete measurements. Deep neural networks, known for their efficiency in handling big data, are anticipated to be adept at extracting underlying information from images sharing certain characteristics, such as specific types of natural or medical images. We introduce a \ac{NFR} method designed to reconstruct images from incomplete and noisy measurements. The method involves training a normalizing flow network to understand the statistical distribution of sample images by mapping them to Gaussian distributions. This well-trained network then acts as a regularization tool within a Bayesian inversion framework. Additionally, we explore the concept of adaptive regularization selection, providing theoretical proof of its admissibility. A significant challenge in three-dimensional image training is the extensive memory and computation requirements. We address this by training the normalizing flow model using only small-size images and applying a patch-based model for reconstructing larger images. Our approach is model-independent, allowing the reuse of a well-trained network as regularization for various imaging systems. Moreover, as a data-driven prior, \ac{NFR} effectively leverages the available dataset information, outperforming artificial priors. This advantage is demonstrated through numerical simulations of three-dimensional photoacoustic tomography under various conditions of sparsity, noise levels, and limited-view scenarios.

Versatile Teacher: A class-aware teacher–student framework for cross-domain adaptation

Addressing the challenge of domain shift between datasets is vital in maintaining model performance. In the context of cross-domain object detection, the teacher–student framework, a widely-used semi-supervised model, has shown significant accuracy improvements. However, existing methods often overlook class differences, treating all classes equally, resulting in suboptimal results. Furthermore, the integration of instance-level alignment with a one-stage detector, essential due to the absence of a Region Proposal Network (RPN), remains unexplored in this framework. In response to these shortcomings, we introduce a novel teacher–student model named Versatile Teacher (VT). VT differs from previous works by considering class-specific detection difficulty and employing a two-step pseudo-label selection mechanism, referred to as Class-aware Pseudo-label Adaptive Selection (CAPS), to generate more reliable pseudo labels. These labels are leveraged as saliency matrices to guide the discriminator for targeted instance-level alignment. Our method demonstrates promising results on three benchmark datasets, and extends the alignment methods for widely-used one-stage detectors, presenting significant potential for practical applications. Code is available at https://github.com/RicardooYoung/VersatileTeacher.

Modulating color cue effectiveness: The role of active selection in visuomotor adaptation

Modulating color cue effectiveness: The role of active selection in visuomotor adaptation

Co‐occurrence in snake and lizard sister species is unrelated to major ecomorphological traits

AbstractEnvironmental factors and interspecific interactions, such as competition and facilitation, can shape species' geographic ranges. Here, we tested the relationship between geographic range overlap of squamate sister species, and their divergence in body size, diel activity, and microhabitat use. Competition theory predicts that sister species with similar traits will overlap less geographically than sister pairs with dissimilar traits. However, similar distributions may present similar selective pressures and favour similar adaptations, such that habitat filtering may result in species with more similar traits in sympatry. Across 1434 sister species contrasts, we found little relationship between range overlap and niche divergence. In some models, the divergence in body size and microhabitat use marginally increased with sympatry, while in other models, sympatric sisters had similar activity times. However, the low R‐squared values of almost all these models lend only weak support to predictions from competition or habitat filtering theories. Sympatric sister species within the same biome showed more similar activity times than expected, lending some support to habitat filtering. Niche divergence in allopatry or sympatry, as calculated using a multi‐trait dissimilarity index, did not show a phylogenetic signal, and niches of sister species from different squamate clades did not deviate significantly from the null expectation. Overall, niche divergence or convergence, across the axes we explored, is not a prerequisite for regional species co‐occurrence. We suggest here that the lack of consistent support for either limiting similarity or habitat filtering may reflect that both these forces act as transient phenomena. More fine‐grained analyses, in space and time, would be needed to detect their fingerprints. Thus, coexistence in the face of competition can arise due to various evolutionary and biogeographic mechanisms, acting concurrently or asynchronously.

An adaptive testing item selection strategy via a deep reinforcement learning approach.

Computerized adaptive testing (CAT) aims to present items that statistically optimize the assessment process by considering the examinee's responses and estimated trait levels. Recent developments in reinforcement learning and deep neural networks provide CAT with the potential to select items that utilize more information across all the items on the remaining tests, rather than just focusing on the next several items to be selected. In this study, we reformulate CAT under the reinforcement learning framework and propose a new item selection strategy based on the deep Q-network (DQN) method. Through simulated and empirical studies, we demonstrate how to monitor the training process to obtain the optimal Q-networks, and we compare the accuracy of the DQN-based item selection strategy with that of five traditional strategies-maximum Fisher information, Fisher information weighted by likelihood, Kullback‒Leibler information weighted by likelihood, maximum posterior weighted information, and maximum expected information-on both simulated and real item banks and responses. We further investigate how sample size and the distribution of the trait levels of the examinees used in training affect DQN performance. The results show that DQN achieves lower RMSE and MAE values than traditional strategies under simulated and real banks and responses in most conditions. Suggestions for the use of DQN-based strategies are provided, as well as their code.

DDPG-AdaptConfig: A deep reinforcement learning framework for adaptive device selection and training configuration in heterogeneity federated learning

Federated Learning (FL) is a distributed machine learning approach that protects user privacy by collaboratively training shared models across devices without sharing their raw personal data. Despite its advantages, FL faces issues of increased convergence time and decreased accuracy due to the heterogeneity of data and systems across devices. Existing methods for solving these issues using reinforcement learning often ignore the adaptive configuration of local training hyperparameters to suit varying data characteristics and system resources. Moreover, they frequently overlook the heterogeneous information contained within local model parameters. To address these problems, we propose the DDPG-AdaptConfig framework based on Deep Deterministic Policy Gradient (DDPG) for adaptive device selection and local training hyperparameters configuration in FL to speed up convergence and ensure high model accuracy. Additionally, we develop a new actor network that integrates the transformer mechanism to extract heterogeneous information from model parameters, which assists in device selection and hyperparameters configuration. Furthermore, we introduce a clustering-based aggregation strategy to accommodate heterogeneity and prevent performance declines. Experimental results show that our DDPG-AdaptConfig achieves significant improvements over existing baselines.

Genetic analysis of tomato brown rugose fruit virus reveals evolutionary adaptation and codon usage bias patterns

Tomato brown rugose fruit virus (ToBRFV) poses a significant threat to tomato production worldwide, prompting extensive research into its genetic diversity, evolutionary dynamics, and adaptive strategies. In this study, we conducted a comprehensive analysis of ToBRFV at the codon level, focusing on codon usage bias, selection pressures, and evolutionary patterns across multiple genes. Our analysis revealed distinct patterns of codon usage bias and selection pressures within the ToBRFV genome, with varying levels of genetic diversity and evolutionary constraints among different genes. We observed a transition/transversion bias of 2.07 across the entire ToBRFV genome, with the movement protein (MP) gene exhibiting the highest transition/transversion bias and SNP density, suggesting potential evolutionary pressures or a higher mutation rate in this gene. Furthermore, our study identified episodic positive selection primarily in the MP gene, highlighting specific codons subject to adaptive changes in response to host immune pressures or environmental factors. Comparative analysis of codon usage bias in the coat protein (CP) and RNA-dependent RNA polymerase (RdRp) genes revealed gene-specific patterns reflecting functional constraints and adaptation to the host's translational machinery. Our findings provide valuable insights into the molecular mechanisms driving ToBRFV evolution and adaptation, with implications for understanding viral pathogenesis, host-virus interactions, and the development of control strategies. Future research directions include further elucidating the functional significance of codon usage biases, exploring the role of episodic positive selection in viral adaptation, and leveraging these insights to inform the development of effective antiviral strategies and crop protection measures.

Use of AI in Pediatric Occupational Therapy: A Review

The utilization of artificial intelligence (AI) in pediatric occupational therapy (OT) has emerged as a promising avenue for enhancing assessment, intervention, and outcomes for children with diverse developmental needs. This paper provides a comprehensive review of the current state of AI applications in pediatric OT, highlighting key findings, benefits, challenges, and future directions. AI technologies, including machine learning algorithms, computer vision systems, and wearable sensors, offer innovative approaches to assess children's motor skills, sensory responses, and cognitive functions objectively and efficiently. AI-driven intervention strategies, such as personalized treatment planning, adaptive task selection, virtual reality environments, and gamified activities, promote engagement, motivation, and skill acquisition among pediatric patients. AI can be helpful in early diagnosis as well as early intervention. Additionally, AI-powered telehealth platforms enable remote delivery of OT services, real-time monitoring of patient progress, and access to care for underserved populations. However, challenges related to data privacy, ethical decision-making, disparities in access, and therapist education must be addressed to ensure the ethical, effective, and equitable integration of AI into pediatric OT practice. By embracing ongoing research, collaboration, and innovation, pediatric OT practitioners can harness the transformative potential of AI to improve outcomes and quality of life for children and families worldwide.

An adaptive operator selection cuckoo search for parameter extraction of photovoltaic models

Accurate, reliable, and efficient extraction of photovoltaic (PV) model parameters is an essential step towards PV system simulation, control, and optimization. Nevertheless, this problem is still facing great challenges because of its intrinsic nonlinear, multivariate, and multimodal properties. In this paper, a new variant of cuckoo search (CS), adaptive operator selection CS (AOSCS), is advanced for the PV model parameter extraction problems. AOSCS includes two major improvements: (1) an adaptive operator selection mechanism is developed to automatically assign the workloads of exploration and exploitation operators, and (2) the exploration and exploitation operators used in the original CS are modified to promote the exploration capability and reduce the blindness of search, respectively. The performance of AOSCS is firstly validated on CEC 2017 test suite and then it is utilized to solve the parameter extraction problems of five PV models. Moreover, further experiments on two commercial PV modules under distinct irradiance and temperature levels are also conducted to evaluate the practicality of the proposed algorithm. It is manifested that the results yielded by AOSCS are very competitive relative to other parameter extraction approaches. Accordingly, the proposed AOSCS is able to be served as an up-and-coming candidate algorithm for PV model parameter extraction problems.

A Trusted Supervision Paradigm for Autonomous Driving Based on Multimodal Data Authentication

At the current stage of autonomous driving, monitoring the behavior of safety stewards (drivers) is crucial to establishing liability in the event of an accident. However, there is currently no method for the quantitative assessment of safety steward behavior that is trusted by multiple stakeholders. In recent years, deep-learning-based methods can automatically detect abnormal behaviors with surveillance video, and blockchain as a decentralized and tamper-resistant distributed ledger technology is very suitable as a tool for providing evidence when determining liability. In this paper, a trusted supervision paradigm for autonomous driving (TSPAD) based on multimodal data authentication is proposed. Specifically, this paradigm consists of a deep learning model for driving abnormal behavior detection based on key frames adaptive selection and a blockchain system for multimodal data on-chaining and certificate storage. First, the deep-learning-based detection model enables the quantification of abnormal driving behavior and the selection of key frames. Second, the key frame selection and image compression coding balance the trade-off between the amount of information and efficiency in multiparty data sharing. Third, the blockchain-based data encryption sharing strategy ensures supervision and mutual trust among the regulatory authority, the logistic platform, and the enterprise in the driving process.

Multi-Task Collaborative Pre-Training and Adaptive Token Selection: A Unified Framework for Brain Representation Learning.

Structural magnetic resonance imaging (sMRI) reveals the structural organization of the brain. Learning general brain representations from sMRI is an enduring topic in neuroscience. Previous deep learning models neglect that the brain, as the core of cognition, is distinct from other organs whose primary attribute is anatomy. Capturing the high-level representation associated with inter-individual cognitive variability is key to appropriately represent the brain. Given that this cognition-related information is subtle, mixed, and distributed in the brain structure, sMRI-based models need to both capture fine-grained details and understand how they relate to the overall global structure. Additionally, it is also necessary to explicitly express the cognitive information that implicitly embedded in local-global image features. Therefore, we propose MCPATS, a brain representation learning framework that combines Multi-task Collaborative Pre-training (MCP) and Adaptive Token Selection (ATS). First, we develop MCP, including mask-reconstruction to understand global context, distort-restoration to capture fine-grained local details, adversarial learning to integrate features at different granularities, and age-prediction, using age as a surrogate for cognition to explicitly encode cognition-related information from local-global image features. This co-training allows progressive learning of implicit and explicit cognition-related representations. Then, we develop ATS based on mutual attention for downstream use of the learned representation. During fine-tuning, the ATS highlights discriminative features and reduces the impact of irrelevant information. MCPATS was validated on three different public datasets for brain disease diagnosis, outperforming competing methods and achieving accurate diagnosis. Further, we performed detailed analysis to confirm that the MCPATS-learned representation captures cognition-related information.