ABSTRACT
A large number of works have presented the single-cell RNA sequencing (scRNA-seq) to study the diversity and biological functions of cells at the single-cell level. Clustering identifies unknown cell types, which is essential for downstream analysis of scRNA-seq samples. However, the high dimensionality, high noise and pervasive dropout rate of scRNA-seq samples have a significant challenge to the cluster analysis of scRNA-seq samples. Herein, we propose a new adaptive fuzzy clustering model based on the denoising autoencoder and self-attention mechanism called the scDASFK. It implements the comparative learning to integrate cell similar information into the clustering method and uses a deep denoising network module to denoise the data. scDASFK consists of a self-attention mechanism for further denoising where an adaptive clustering optimization function for iterative clustering is implemented. In order to make the denoised latent features better reflect the cell structure, we introduce a new adaptive feedback mechanism to supervise the denoising process through the clustering results. Experiments on 16 real scRNA-seq datasets show that scDASFK performs well in terms of clustering accuracy, scalability and stability. Overall, scDASFK is an effective clustering model with great potential for scRNA-seq samples analysis. Our scDASFK model codes are freely available at https://github.com/LRX2022/scDASFK.
Subject(s)
Gene Expression Profiling , Single-Cell Analysis , Gene Expression Profiling/methods , Sequence Analysis, RNA/methods , Single-Cell Analysis/methods , Cluster Analysis , AlgorithmsABSTRACT
N6-methyladinosine (m6A) modification is the most abundant co-transcriptional modification in eukaryotic RNA and plays important roles in cellular regulation. Traditional high-throughput sequencing experiments used to explore functional mechanisms are time-consuming and labor-intensive, and most of the proposed methods focused on limited species types. To further understand the relevant biological mechanisms among different species with the same RNA modification, it is necessary to develop a computational scheme that can be applied to different species. To achieve this, we proposed an attention-based deep learning method, adaptive-m6A, which consists of convolutional neural network, bi-directional long short-term memory and an attention mechanism, to identify m6A sites in multiple species. In addition, three conventional machine learning (ML) methods, including support vector machine, random forest and logistic regression classifiers, were considered in this work. In addition to the performance of ML methods for multi-species prediction, the optimal performance of adaptive-m6A yielded an accuracy of 0.9832 and the area under the receiver operating characteristic curve of 0.98. Moreover, the motif analysis and cross-validation among different species were conducted to test the robustness of one model towards multiple species, which helped improve our understanding about the sequence characteristics and biological functions of RNA modifications in different species.
Subject(s)
Machine Learning , RNA , Base Sequence , RNA/genetics , Neural Networks, ComputerABSTRACT
A novel advantage of the use of machine learning (ML) systems in medicine is their potential to continue learning from new data after implementation in clinical practice. To date, considerations of the ethical questions raised by the design and use of adaptive machine learning systems in medicine have, for the most part, been confined to discussion of the so-called "update problem," which concerns how regulators should approach systems whose performance and parameters continue to change even after they have received regulatory approval. In this paper, we draw attention to a prior ethical question: whether the continuous learning that will occur in such systems after their initial deployment should be classified, and regulated, as medical research? We argue that there is a strong prima facie case that the use of continuous learning in medical ML systems should be categorized, and regulated, as research and that individuals whose treatment involves such systems should be treated as research subjects.
ABSTRACT
A wide variety of weather conditions, from windstorms to prolonged heat events, can substantially impact power systems, posing many risks and inconveniences due to power outages. Accurately estimating the probability distribution of the number of customers without power using data about the power utility system and environmental and weather conditions can help utilities restore power more quickly and efficiently. However, the critical shortcoming of current models lies in the difficulties of handling (i) data streams and (ii) model uncertainty due to combining data from various weather events. Accordingly, this article proposes an adaptive ensemble learning algorithm for data streams, which deploys a feature- and performance-based weighting mechanism to adaptively combine outputs from multiple competitive base learners. As a proof of concept, we use a large, real data set of daily customer interruptions to develop the first adaptive all-weather outage prediction model using data streams. We benchmark several approaches to demonstrate the advantage of our approach in offering more accurate probabilistic predictions. The results show that the proposed algorithm reduces the probabilistic predictions' error of the base learners between 4% and 22% with an average of 8%, which also result in substantially more accurate point predictions. The improvement made by our algorithm is enhanced as we exchange base learners with simpler models.
ABSTRACT
Direct policy search (DPS) is a method for identifying optimal policies (i.e., rules) for managing a system in response to changing conditions. In this article, we introduce a new adaptive way to incorporate learning into DPS. The standard DPS approach identifies "robust" policies by optimizing their average performance over a large ensemble of future states of the world (SOW). Our approach exploits information gained over time, updating prior beliefs about the kind of SOW being experienced. We first run the standard DPS approach multiple times, but with varying sets of weights applied to the SOWs when calculating average performance. Adaptive "metapolicies" then further improve performance by specifying how control of the system should switch between policies identified using different weight sets, depending on our updated beliefs about the relative likelihood of being in certain SOWs. We outline the general method and illustrate it using a case study of efficient dike heightening that simultaneously minimizes protection system costs and flood damage resulting from rising sea levels and storm surge. The solutions identified by our adaptive algorithm dominate the standard DPS on these two objectives, with an average marginal damage reduction of 35.1% for policies with similar costs; improvements are largest in SOWs with relatively lower sea level rise. We also evaluate how performance varies under different ways of implementing the algorithm, such as changing the frequency with which beliefs are updated.
ABSTRACT
The United States military services are modernizing their training and education curricula by leveraging advances in technology to deliver instruction that is more engaging and responsive to trainees' needs and better prepares them for the future fight. Adaptive training (AT), or training tailored to the strengths and weaknesses of individual trainees, is a promising technique to meet these modernization goals. The research literature, however, is sporadic and does not clearly prescribe best practices for its employment. Therefore, we conducted a meta-analysis to examine the effectiveness of various AT instructional interventions (i.e. adapting difficulty, feedback, scaffolding, etc.) on learning outcomes. There were 30 peer-reviewed publications included in the analysis. We grouped studies by the adaptive intervention examined and reported the associated effects on learning outcomes. Overall, the results revealed that the effectiveness of AT varied considerably across the instructional interventions. Specifically, studies that implemented adaptive difficulty techniques were the most effective, followed by adaptive scaffolding and remediation/test-out techniques. Based on these findings, we identify design recommendations for future AT systems.
ABSTRACT
People adjust their learning rate rationally according to local environmental statistics and calibrate such adjustments based on the broader statistical context. To date, no theory has captured the observed range of adaptive learning behaviors or the complexity of its neural correlates. Here, we attempt to do so using a neural network model that learns to map an internal context representation onto a behavioral response via supervised learning. The network shifts its internal context on receiving supervised signals that are mismatched to its output, thereby changing the "state" to which feedback is associated. A key feature of the model is that such state transitions can either increase learning or decrease learning depending on the duration over which the new state is maintained. Sustained state transitions that occur after changepoints facilitate faster learning and mimic network reset phenomena observed in the brain during rapid learning. In contrast, state transitions after one-off outlier events are short lived, thereby limiting the impact of outlying observations on future behavior. State transitions in our model provide the first mechanistic interpretation for bidirectional learning signals, such as the P300, that relate to learning differentially according to the source of surprising events and may also shed light on discrepant observations regarding the relationship between transient pupil dilations and learning. Together, our results demonstrate that dynamic latent state representations can afford normative inference and provide a coherent framework for understanding neural signatures of adaptive learning across different statistical environments.SIGNIFICANCE STATEMENT How humans adjust their sensitivity to new information in a changing world has remained largely an open question. Bridging insights from normative accounts of adaptive learning and theories of latent state representation, here we propose a feedforward neural network model that adjusts its learning rate online by controlling the speed of transitioning its internal state representations. Our model proposes a mechanistic framework for explaining learning under different statistical contexts, explains previously observed behavior and brain signals, and makes testable predictions for future experimental studies.
Subject(s)
Brain , Neural Networks, Computer , Brain/physiology , Feedback , HumansABSTRACT
Accumulating evidence has shown that microRNAs (miRNAs) play crucial roles in different biological processes, and their mutations and dysregulations have been proved to contribute to tumorigenesis. In silico identification of disease-associated miRNAs is a cost-effective strategy to discover those most promising biomarkers for disease diagnosis and treatment. The increasing available omics data sources provide unprecedented opportunities to decipher the underlying relationships between miRNAs and diseases by computational models. However, most existing methods are biased towards a single representation of miRNAs or diseases and are also not capable of discovering unobserved associations for new miRNAs or diseases without association information. In this study, we present a novel computational method with adaptive multi-source multi-view latent feature learning (M2LFL) to infer potential disease-associated miRNAs. First, we adopt multiple data sources to obtain similarity profiles and capture different latent features according to the geometric characteristic of miRNA and disease spaces. Then, the multi-modal latent features are projected to a common subspace to discover unobserved miRNA-disease associations in both miRNA and disease views, and an adaptive joint graph regularization term is developed to preserve the intrinsic manifold structures of multiple similarity profiles. Meanwhile, the Lp,q-norms are imposed into the projection matrices to ensure the sparsity and improve interpretability. The experimental results confirm the superior performance of our proposed method in screening reliable candidate disease miRNAs, which suggests that M2LFL could be an efficient tool to discover diagnostic biomarkers for guiding laborious clinical trials.
Subject(s)
Computational Biology/methods , MicroRNAs/genetics , Biomarkers/metabolism , Carcinoma, Hepatocellular/genetics , Carcinoma, Renal Cell/genetics , Computer Simulation , Humans , Kidney Neoplasms/genetics , Liver Neoplasms/geneticsABSTRACT
In recent times, much-coveted memristor emulators have found their use in a variety of applications such as neuromorphic computing, analog computations, signal processing, etc. Thus, a 100 MHz flux-controlled memristor emulator is proposed in this research brief. The proposed memristor emulator is designed using a single differential voltage current conveyor (DVCC), three PMOS transistors, and one capacitor. Among three PMOS transistors, two transistors are used to implement an active resistor, and one transistor is used as the multiplier required for the necessary memristive behaviors. Through simple adjustment of the switch, the proposed emulator can be operated in incremental as well as decremental configurations. The simulations are performed using a 180 nm technology node to validate the proposed design and are experimentally verified using AD844AN and CD4007 ICs. The memristor states of the proposed emulator are perfectly retained even in the absence of external stimuli, thereby ascertaining the non-volatility behavior. The robustness of the design is further analyzed using the PVT and Monte Carlo simulations, which suggest that the circuit operation is not hindered by the mismatch and process variations. A simple neuromorphic adaptive learning circuit based on the proposed memristor is also designed as an application.
ABSTRACT
Ship trajectory classification is of great significance for shipping analysis and marine security governance. However, in order to cover up their illegal fishing or espionage activities, some illicit ships will forge the ship type information in the Automatic Identification System (AIS), and this label noise will significantly impact the algorithm's classification accuracy. Sample selection is a common and effective approach in the field of learning from noisy labels. However, most of the existing methods based on sample selection need to determine the noise rate of the data through prior means. To address these issues, we propose a noise rate adaptive learning mechanism that operates without prior conditions. This mechanism is integrated with the robust training paradigm JoCoR (joint training with co-regularization), giving rise to a noise rate adaptive learning robust training paradigm called A-JoCoR. Experimental results on real-world trajectories provided by the Danish Maritime Authority verified the effectiveness of A-JoCoR. It not only realizes the adaptive learning of the data noise rate during the training process, but also significantly improves the classification performance compared with the original method.
ABSTRACT
Online learning systems are able to offer customized content catered to individual learner's needs, and have seen growing interest from industry and academia alike in recent years. In contrast to the traditional computerized adaptive testing setting, which has a well-calibrated item bank with new items added periodically, the online learning system has two unique features: (1) the number of items is large, and they have likely not gone through costly field testing for item calibration; and (2) the individual's ability may change as a result of learning. The Elo rating system has been recognized as an effective method for fast updating of item and person parameters in online learning systems to enable personalized learning. However, the updating parameter in Elo has to be tuned post hoc, and Elo is only suitable for the Rasch model. In this paper, we propose the use of a moment-matching Bayesian update algorithm to estimate item and person parameters on the fly. With sequentially updated item and person parameters, a modified maximum posterior weighted information criterion (MPWI) is proposed to adaptively assign items to individuals. The Bayesian updated algorithm along with MPWI is validated in a simulated multiple-session online learning setting, and the results show that the new combo can achieve fast and reasonably accurate parameter estimations that are comparable to random selection, match-difficulty selection, and traditional online calibration. Moreover, the combo can still function reasonably well with as low as 20% of items being pre-calibrated in the item bank.
Subject(s)
Algorithms , Education, Distance , Humans , Bayes Theorem , Calibration , Online Systems , Psychometrics/methodsABSTRACT
OBJECTIVE: Restrictions on working time and healthcare expenditures, as well as increasing subspecialization with caseload requirements per surgeon and increased quality-of-care expectations, provide limited opportunities for surgical residents to be trained in the operating room. Yet, surgical training requires goal-oriented and focused practice. As a result, training simulators are increasingly utilized. The authors designed a two-step blended course consisting of a personalized adaptive electronic learning (e-learning) module followed by simulator training. This paper reports on course development and the evaluation by the first participants. METHODS: Adaptive e-learning was curated by learning engineers based on theoretical information provided by clinicians (subject matter experts). A lumbar spine model for image-guided spinal injections was used for the simulator training. Residents were assigned to the e-learning module first; after its completion, they participated in the simulator training. Performance data were recorded for each participant's e-learning module, which was necessary to personalize the learning experience to each individual's knowledge and needs. Simulator training was organized in small groups with a 1-to-4 instructor-to-participant ratio. Structured assessments were undertaken, adapted from the Student Evaluation of Educational Quality. RESULTS: The adaptive e-learning module was curated, reviewed, and approved within 10 weeks. Eight participants have taken the course to date. The overall rating of the course is very good (4.8/5). Adaptive e-learning is well received compared with other e-learning types (8/10), but scores lower regarding usefulness, efficiency, and fun compared with the simulator training, despite improved conscious competency (32.6% ± 15.1%) and decreased subconscious incompetency (22.8% ± 10.2%). The subjective skill level improved by 20%. Asked about the estimated impact of the course, participants indicated that they had either learned something new that they plan to use in their practice (71.4%) or felt reassured in their practice (28.6%). CONCLUSIONS: The development of a blended training course combining adaptive e-learning and simulator training in a rapid manner is feasible and leads to improved skills. Simulator training is rated more valuable by surgical trainees than theoretical e-learning; the impact of this type of training on patient care needs to be further investigated.
Subject(s)
Internship and Residency , Neurosurgery , Clinical Competence , Curriculum , Humans , Neurosurgery/education , Neurosurgical Procedures/educationABSTRACT
BACKGROUND: Learning to adapt to changes in the environment is highly beneficial. This is especially true for echolocating bats that forage in diverse environments, moving between open spaces to highly complex ones. Bats are known for their ability to rapidly adjust their sensing according to auditory information gathered from the environment within milliseconds but can they also benefit from longer adaptive processes? In this study, we examined adult bats' ability to slowly adapt their sensing strategy to a new type of environment they have never experienced for such long durations, and to then maintain this learned echolocation strategy over time. RESULTS: We show that over a period of weeks, Pipistrellus kuhlii bats gradually adapt their pre-takeoff echolocation sequence when moved to a constantly cluttered environment. After adopting this improved strategy, the bats retained an ability to instantaneously use it when placed back in a similarly cluttered environment, even after spending many months in a significantly less cluttered environment. CONCLUSIONS: We demonstrate long-term adaptive flexibility in sensory acquisition in adult animals. Our study also gives further insight into the importance of sensory planning in the initiation of a precise sensorimotor behavior such as approaching for landing.
Subject(s)
Adaptation, Physiological , Chiroptera , Echolocation , Animals , Flight, AnimalABSTRACT
The pedestrian stride-length estimation is a crucial piece of personal behavior data for many smartphone applications, such as health monitoring and indoor location. The performance of the present stride-length algorithms is suitable for simple gaits and single scenes, but when applied to sophisticated gaits or heterogeneous devices, their inaccuracy varies dramatically. This paper proposes an efficient learning-based stride-length estimation model using a smartphone to obtain the correct stride length. The model uses adaptive learning to extract different elements for changing and recognition tasks, including Long Short-Term Memory (LSTM) and Convolutional Neural Network (CNN) modules. The direct fusion method maps the eigenvectors to the appropriate stride length after combining the features from the learning modules. We presented an online learning module to update the model to increase the SLE model's generalization. Extensive experiments are conducted with heterogeneous devices or users, various gaits, and switched scenarios. The results confirm that the proposed method outperforms other state-of-the-art methods and achieves an average 4.26% estimation error rate in various environments.
Subject(s)
Gait , Pedestrians , Humans , Algorithms , Neural Networks, Computer , SmartphoneABSTRACT
Adaptive systems and Augmented Reality are among the most promising technologies in teaching and learning processes, as they can be an effective tool for training engineering students' spatial skills. Prior work has investigated the integration of AR technology in engineering education, and more specifically, in spatial ability training. However, the modeling of user knowledge in order to personalize the training has been neither sufficiently explored nor exploited in this task. There is a lot of space for research in this area. In this work, we introduce a novel personalization of the learning path within an AR spatial ability training application. The aim of the research is the integration of Augmented Reality, specifically in engineering evaluation and fuzzy logic technology. During one academic semester, three engineering undergraduate courses related to the domain of spatial skills were supported by a developed adaptive training system named PARSAT. Using the technology of fuzzy weights in a rule-based decision-making module and the learning theory of the Structure of the Observed Learning Outcomes for the design of the learning material, PARSAT offers adaptive learning activities for the students' cognitive skills. Students' data were gathered at the end of the academic semester, and a thorough analysis was delivered. The findings demonstrated that the proposed training method outperformed the traditional method that lacked adaptability, in terms of domain expertise and learning theories, considerably enhancing student learning outcomes.
Subject(s)
Augmented Reality , Spatial Navigation , Humans , Knowledge , Learning , StudentsABSTRACT
Innovation in wireless communications and microtechnology has progressed day by day, and this has resulted in the creation of wireless sensor networks. This technology is utilised in a variety of settings, including battlefield surveillance, home security, and healthcare monitoring, among others. However, since tiny batteries with very little power are used, this technology has power and target monitoring issues. With the development of various architectures and algorithms, considerable research has been done to address these problems. The adaptive learning automata algorithm (ALAA) is a scheduling machine learning method that is utilised in this study. It offers a time-saving scheduling method. As a result, each sensor node in the network has been outfitted with learning automata, allowing them to choose their appropriate state at any given moment. The sensor is in one of two states: active or sleep. Several experiments were conducted to get the findings of the suggested method. Different parameters are utilised in this experiment to verify the consistency of the method for scheduling the sensor node so that it can cover all of the targets while using less power. The experimental findings indicate that the proposed method is an effective approach to schedule sensor nodes to monitor all targets while using less electricity. Finally, we have benchmarked our technique against the LADSC scheduling algorithm. All of the experimental data collected thus far demonstrate that the suggested method has justified the problem description and achieved the project's aim. Thus, while constructing an actual sensor network, our suggested algorithm may be utilised as a useful technique for scheduling sensor nodes.
Subject(s)
Computer Communication Networks , Wireless Technology , Algorithms , Machine Learning , Monitoring, PhysiologicABSTRACT
Aiming at the problems of target model drift or loss of target tracking caused by serious deformation, occlusion, fast motion, and out of view of the target in long-term moving target tracking in complex scenes, this paper presents a robust multi-feature single-target tracking algorithm based on a particle filter. The algorithm is based on the correlation filtering framework. First, to extract more accurate target appearance features, in addition to the manual features histogram of oriented gradient features and color histogram features, the depth features from the conv3-4, conv4-4 and conv5-4 convolutional layer outputs in VGGNet-19 are also fused. Secondly, this paper designs a re-detection module of a fusion particle filter for the problem of how to return to accurate tracking after the target tracking fails, so that the algorithm in this paper can maintain high robustness during long-term tracking. Finally, in the adaptive model update stage, the adaptive learning rate update and adaptive filter update are performed to improve the accuracy of target tracking. Extensive experiments are conducted on dataset OTB-2015, dataset OTB-2013, and dataset UAV123. The experimental results show that the proposed multi-feature single-target robust tracking algorithm with fused particle filtering can effectively solve the long-time target tracking problem in complex scenes, while showing more stable and accurate tracking performance.
ABSTRACT
Over the past couple of decades, many telecommunication industries have passed through the different facets of the digital revolution by integrating artificial intelligence (AI) techniques into the way they run and define their processes. Relevant data acquisition, analysis, harnessing, and mining are now fully considered vital drivers for business growth in these industries. Machine learning, a subset of artificial intelligence (AI), can assist, particularly in learning patterns in big data chunks, intelligent extrapolative extraction of data and automatic decision-making in predictive learning. Firstly, in this paper, a detailed performance benchmarking of adaptive learning capacities of different key machine-learning-based regression models is provided for extrapolative analysis of throughput data acquired at the different user communication distances to the gNodeB transmitter in 5G new radio networks. Secondly, a random forest (RF)-based machine learning model combined with a least-squares boosting algorithm and Bayesian hyperparameter tuning method for further extrapolative analysis of the acquired throughput data is proposed. The proposed model is herein referred to as the RF-LS-BPT method. While the least-squares boosting algorithm is engaged to turn the possible RF weak learners to form stronger ones, resulting in a single strong prediction model, the Bayesian hyperparameter tuning automatically determines the best RF hyperparameter values, thereby enabling the proposed RF-LS-BPT model to obtain desired optimal prediction performance. The application of the proposed RF-LS-BPT method showed superior prediction accuracy over the ordinary random forest model and six other machine-learning-based regression models on the acquired throughput data. The coefficient of determination (Rsq) and mean absolute error (MAE) values obtained for the throughput prediction at different user locations using the proposed RF-LS-BPT method range from 0.9800 to 0.9999 and 0.42 to 4.24, respectively. The standard RF models attained 0.9644 to 0.9944 Rsq and 5.47 to 12.56 MAE values. The improved throughput prediction accuracy of the proposed RF-LS-BPT method demonstrates the significance of hyperparameter tuning/optimization in developing precise and reliable machine-learning-based regression models. The projected model would find valuable applications in throughput estimation and modeling in 5G and beyond 5G wireless communication systems.
Subject(s)
Artificial Intelligence , Machine Learning , Algorithms , Bayes Theorem , Least-Squares AnalysisABSTRACT
Contemporary education is often based on using e-learning courses, which have become a popular means of delivering didactic material to students. Among the main advantages mentioned is the potential possibility of creating individual ways of learning and teaching. The purpose of this article is to provide a description of the various approaches to adaptive learning, to present the comparative research results of a Polish-Ukrainian study, as well as to highlight the options offered by LMS Moodle (Modular object-oriented dynamic learning environment Learning Management System) for the implementation of adaptive learning and the possibility of taking into account the expectation of students regardless of the country. During the study, an online survey was administered to 59 students at the University of Silesia in Katowice (US), Poland and 121 students at the Borys Grinchenko Kyiv University (BGKU), Ukraine between March-June 2020. All of the students were studying online due to the pandemic caused by COVID-19. The classes were conducted in synchronous and in asynchronous modes using e-learning courses in the Moodle system, as well as MS Teams for Polish students and Google Suite for Ukrainian students. At the same time, the majority of the surveyed students declared that they lack personalization, both in terms of materials and the learning process, which was limited in terms of fulfilment and they would like to have a choice of the level of study.
ABSTRACT
This is a case study on precision education (PE) in a school using the Taiwan adaptive learning platform (TALP), established by the Ministry of Education of Taiwan. TALP can be viewed as a form of PE because it can identify students' learning deficits, offer various learning materials, and provide feedback. There are limited studies on PE. The ways in which TALP enhances teaching and learning is meaningful for educational technologies, and to help improve TALP and applications of PE. Three fifth-grade classes in a Taiwanese elementary school were selected to participate in this study, comprising a total of 76 students and two teachers. Pre/post tests and regular tests were conducted. The quantitative data were analyzed with nonparametric statistics by using the Kruskal-Wallis Test and the Median Test, which did not show significant differences. However, based on the qualitative data, teachers confirmed PE could help them identify students' learning deficits. Further, this study suggests that TALP and PE may enhance students' learning to identify students' learning challenges and offer feedback. Future research should focus on a longitudinal methodology and additional PE application studies; furthermore, research should also collect students' observations and interviews.