Training Strategies Research Articles

A residual GRU method with deep cross fusion for Alzheimer’s disease progression prediction using missing variable-length time series data

Alzheimer’s disease (AD) is a neurodegenerative disorder with a long prodromal phase. Early prediction of AD progression is crucial for improving clinical diagnosis. However, missing data and variable-length of time series data make it difficult to predict the progression of AD in clinical practice. Regarding this problem, existing researches mainly employed LSTM (Long Short-Term Memory) networks to impute missing data and make predictions, but those methods only encoded fixed-length time series data and suffered from the error accumulation, resulting in relatively unsatisfactory prediction results. To address this issue, this paper develops a novel method, named residual sharing GRU (Gated Recurrent Unit) with enhanced deep cross fusion module (RGRU-EDCF), to perform AD progression prediction. Specifically, we first design an enhanced cross deep module to impute missing data and learn complementary information. Moreover, we design a residual sharing GRU model to realize the input of variable-length time series data, thus, enhancing flexibility of model. Besides, the residual structure in the residual sharing GRU model is used for reducing error accumulation. We also add a local multi-head attention mechanism for specific time series features learning for classification and regression prediction. Finally, we use specific time series features to realize prediction of classification and regression tasks. Unlike existing methods that trained the missing data imputation and prediction model separately, our method considers an end-to-end training strategy to train both enhanced cross deep module and the residual sharing GRU model to further promote the performance of AD progression prediction. The proposed method is validated using time series data from the ADNI dataset. The accuracy can reach 95.17% in the prediction task, MAE can reach 2.64 in cognitive score prediction, and 4.04 in MRI biomarkers prediction, it is competitive and superior over other state-of-the-art methods. We also validate our proposed method on MIMIC-III dataset.

Diffraction model-driven neural network with semi-supervised training strategy for real-world 3D holographic photography

Compared to traditional 2D displays, 3D display technology provides richer information to the viewer. Learning-based computer-generated holography (CGH) has shown great potential in realizing real-time holographic 3D displays. However, most of the current learning-based CGH algorithms cannot quickly complete the training stage and produce high-quality holograms due to insufficient constraints in the training stage of the neural network. In this paper, we propose a diffractive model-driven neural network trained using a semi-supervised training (SST-holo) strategy and incorporate a state-of-the-art monocular depth estimation algorithm to achieve the fast generation of holograms of real-world 3D scenes. Compared to the supervised training strategy, our proposed semi-supervised training strategy does not require high-quality labeled datasets, but can significantly improve the imaging quality and generalization of the algorithm. Incorporating the Res-MSR block in SST-holo to adaptively learn image features of different scales enhances the learning capability of the network. In addition, we adopt a random splicing processing strategy to preprocess the dataset to ensure that the original features in the dataset are not corrupted. SST-holo can generate high-quality 3D phase-only holograms with 2 K resolution in 0.015 seconds. Both monochrome and color optical experiments show that the proposed algorithm has good 3D effect and generalization ability and can effectively improve the quality of reconstructed images.

Machine learning predictive model to estimate the photo-degradation performance of stannates and hydroxystannates photocatalysts on a variety of waterborne contaminants

In this work, a comprehensive machine learning (ML) methodology was used to predict the degradation efficiency of different stannate and hydroxystannate photocatalysts on a wide range of waterborne pollutants. The structural, atomic features along with molecular fingerprints (MF) were used as descriptors of the crystalline phase of the photocatalysts and the organic compounds, respectively. The encoded features of the photocatalysts and contaminants along with the experimental variables of the degradation process are input to two ML models, named as RF (random forest) and KNN (K nearest neighbor). The RF model has achieved a very good prediction of the photocatalytic degradation efficiency (%) by different photocatalysts over a wide range of organic contaminants. The RF model performance was investigated by applying two different training strategies. The effects of different factors on photocatalytic degradation performance are further evaluated by feature importance analyses. Two illustrative applications on the use of the ML model for optimal photocatalyst selection and for assessing other types of photocatalysts for different environmental applications were provided.

Advancing Real-World Image Dehazing: Perspective, Modules, and Training.

Restoring high-quality images from degraded hazy observations is a fundamental and essential task in the field of computer vision. While deep models have achieved significant success with synthetic data, their effectiveness in real-world scenarios remains uncertain. To improve adaptability in real-world environments, we construct an entirely new computational framework by making efforts from three key aspects: imaging perspective, structural modules, and training strategies. To simulate the often-overlooked multiple degradation attributes found in real-world hazy images, we develop a new hazy imaging model that encapsulates multiple degraded factors, assisting in bridging the domain gap between synthetic and real-world image spaces. In contrast to existing approaches that primarily address the inverse imaging process, we design a new dehazing network following the "localization-and-removal" pipeline. The degradation localization module aims to assist in network capture discriminative haze-related feature information, and the degradation removal module focuses on eliminating dependencies between features by learning a weighting matrix of training samples, thereby avoiding spurious correlations of extracted features in existing deep methods. We also define a new Gaussian perceptual contrastive loss to further constrain the network to update in the direction of the natural dehazing. Regarding multiple full/no-reference image quality indicators and subjective visual effects on challenging RTTS, URHI, and Fattal real hazy datasets, the proposed method has superior performance and is better than the current state-of-the-art methods.

Rapid cycle deliberate practice for an ECMO emergency simulation training for nurses: A quasi-experimental study

BackgroundThe air entering the ECMO circuit is considered a mechanical emergency with a high mortality rate. Team training for nurses is necessary to handle these emergencies. MethodQuasi-experimental pre- and post-intervention study. Data were collected from October 2021 to February 2022 at a hospital in Rio de Janeiro, Brazil. The intervention was the Rapid Cycle Deliberate Practice simulation training. The analysis included comparing performance between pre and post-test, time to action, and satisfaction and self-confidence. Inferential statistics and intervention effect size were calculated. Results28 nurses participated in the research. Performance scores increased by 44.1% in the post-test, with an average reduction of 3.54 minutes in time spent on actions (p < .0001). The effect size of the intervention was high, and participants were satisfied and self-confident. ConclusionRCDP can be applied as a simulation strategy for training in the management of an arterial air ECMO emergency, with improved participant performance.

One-Stage Anchor-Free Online Multiple Target Tracking With Deformable Local Attention and Task-Aware Prediction.

The tracking-by-detection paradigm currently dominates multiple target tracking algorithms. It usually includes three tasks: target detection, appearance feature embedding, and data association. Carrying out these three tasks successively usually leads to lower tracking efficiency. In this paper, we propose a one-stage anchor-free multiple task learning framework which carries out target detection and appearance feature embedding in parallel to substantially increase the tracking speed. This framework simultaneously predicts a target detection and produces a feature embedding for each location, by sharing a pyramid of feature maps. We propose a deformable local attention module which utilizes the correlations between features at different locations within a target to obtain more discriminative features. We further propose a task-aware prediction module which utilizes deformable convolutions to select the most suitable locations for the different tasks. At the selected locations, classification of samples into foreground or background, appearance feature embedding, and target box regression are carried out. Two effective training strategies, regression range overlapping and sample reweighting, are proposed to reduce missed detections in dense scenes. Ambiguous samples whose identities are difficult to determine are effectively dealt with to obtain more accurate feature embedding of target appearance. An appearance-enhanced non-maximum suppression is proposed to reduce over-suppression of true targets in crowded scenes. Based on the one-stage anchor-free network with the deformable local attention module and the task-aware prediction module, we implement a new online multiple target tracker. Experimental results show that our tracker achieves a very fast speed while maintaining a high tracking accuracy.

A Phone-Based Distributed Ambient Temperature Measurement System With an Efficient Label-Free Automated Training Strategy

A Phone-Based Distributed Ambient Temperature Measurement System With an Efficient Label-Free Automated Training Strategy

MCTformer+: Multi-Class Token Transformer for Weakly Supervised Semantic Segmentation.

This paper proposes a novel transformer-based framework to generate accurate class-specific object localization maps for weakly supervised semantic segmentation (WSSS). Leveraging the insight that the attended regions of the one-class token in the standard vision transformer can generate class-agnostic localization maps, we investigate the transformer's capacity to capture class-specific attention for class-discriminative object localization by learning multiple class tokens. We present the Multi-Class Token transformer, which incorporates multiple class tokens to enable class-aware interactions with patch tokens. This is facilitated by a class-aware training strategy that establishes a one-to-one correspondence between output class tokens and ground-truth class labels. We also introduce a Contrastive-Class-Token (CCT) module to enhance the learning of discriminative class tokens, enabling the model to better capture the unique characteristics of each class. Consequently, the proposed framework effectively generates class-discriminative object localization maps from the class-to-patch attentions associated with different class tokens. To refine these localization maps, we propose the utilization of patch-level pairwise affinity derived from the patch-to-patch transformer attention. Furthermore, the proposed framework seamlessly complements the Class Activation Mapping (CAM) method, yielding significant improvements in WSSS performance on PASCAL VOC 2012 and MS COCO 2014. These results underline the importance of the class token for WSSS.

Learning Many-to-Many Mapping for Unpaired Real-World Image Super-Resolution and Downscaling.

Learning based single image super-resolution (SISR) for real-world images has been an active research topic yet a challenging task, due to the lack of paired low-resolution (LR) and high-resolution (HR) training images. Most of the existing unsupervised real-world SISR methods adopt a two-stage training strategy by synthesizing realistic LR images from their HR counterparts first, then training the super-resolution (SR) models in a supervised manner. However, the training of image degradation and SR models in this strategy are separate, ignoring the inherent mutual dependency between downscaling and its inverse upscaling process. Additionally, the ill-posed nature of image degradation is not fully considered. In this paper, we propose an image downscaling and SR model dubbed as SDFlow, which simultaneously learns a bidirectional many-to-many mapping between real-world LR and HR images unsupervisedly. The main idea of SDFlow is to decouple image content and degradation information in the latent space, where content information distribution of LR and HR images is matched in a common latent space. Degradation information of the LR images and the high-frequency information of the HR images are fitted to an easy-to-sample conditional distribution. Experimental results on real-world image SR datasets indicate that SDFlow can generate diverse realistic LR and SR images both quantitatively and qualitatively.

The impact on users of an unfamiliar AED following a recent training experience: A randomized cross over simulation study

ObjectiveThis study examined the impact of prior familiarity with automated external defibrillator (AED) models on the time of defibrillation and the emotional experiences of laypersons. MethodsWe conducted a randomized cross over simulation study with 123 participants to assess their reactions to both familiar and unfamiliar AED models. The time to first defibrillation was measured using three different AED training models, two of which were previously unknown to the participants. Additionally, semi-structured interviews were held with the participants to gather further insights. ResultsParticipants took longer to initiate defibrillation with unfamiliar (M = 34 s) AEDs compared to familiar (M = 27 s) ones. This delay was accompanied by feelings of confusion, nervousness, and anxiety. Factors such as the design of the AED covers, electrodes, and buttons were identified as sources of confusion. Nonetheless, clear instructions and similarities between devices helped facilitate their use. ConclusionThe findings suggest that AED design and familiarity with different AED designs may affect performance by laypersons. To improve user confidence, it would be useful to familiarize users with a variety of AED models as part of training initiatives. Understanding the impact of AED familiarity on rescuer’s response can guide CPR training strategies and improve outcomes for OHCA. As more AED models become available to the public, the user-friendliness of AEDs may also be improved. It is beneficial for AED manufacturers to consider the results of research when developing new models.

MPE[formula omitted]: Learning meta-prompt with entity-enhanced semantics for few-shot named entity recognition

Recently, prompt-tuning has been proven to be surprisingly effective on few-shot tasks. Intuitively, some studies explore Few-shot Named Entity Recognition (NER) based on prompt-tuning. However, how to properly initialize and effectively learn the prompt under limited training conditions still remains significantly challenging for few-shot NER. To meet these challenges, we propose a novel Meta-Prompt with Entity-Enhanced semantics for Few-shot NER, MPE3 for brevity. Specifically, we first explore the importance of the named entities’ semantics in the few-shot NER task. And we propose to construct prompts with entity-enhanced semantics which contain much useful prior knowledge for identifying named entities. Furthermore, to address the issue of inadequate training more substantially, we aim to train a meta-prompt that can be more effective and adaptive for few-shot NER scenarios. To achieve this, we divide the training data into many source-domain agnostic meta-tasks tailored to the characteristics of the NER problem for training. And we specially design a prompt meta-learner for training these meta-tasks. This training strategy succeeds in guiding prompts to optimize in a better direction for few-shot scenarios by the learned meta-knowledge from each meta-task. We conduct extensive experiments on three NER datasets under two different few-shot settings. Our method outperforms the current state-of-the-art model by 5.60%∼13.34% and 2.41%∼7.34% on average in the two different few-shot settings respectively, which validates the effectiveness and superiority of our model.

Self-supervised anomaly detection and localization for X-ray cargo images: Generalization to novel anomalies

Robust detection of illicit items using X-ray inspection methods has gained increasing importance in recent years due to the large volume of cargo crossing international borders. In addition to detecting the presence of such items, determining their location, size, and shape is challenging due to the unpredictable nature of anomalies, but essential for expediting security inspections. Viewing the illicit items as anomalies relative to expected cargo, we propose a self-supervised learning framework consisting of an encoder–decoder–classifier–segmenter model, a multi-component loss function, coupled with a training strategy to extract discriminative features tailored for detection of the presence of anomalies, as well as localization of such items in X-ray cargo images. Our framework addresses the challenges posed by limited labeled data and offers a model capable of both detecting and localizing anomalies effectively. Moreover, we present a diverse dataset encompassing various cargo scenarios with and without anomalies, providing a robust evaluation environment for this class of problems. Unlike existing approaches, which are trained to detect specific types of objects with a fixed set of illicit items, our framework is adaptable to real-world scenarios where a wide range of illicit items may be present in the cargo. This versatility enhances the practical applicability of our model. We evaluate the performance of our framework on our dataset as well as two other publicly available datasets, demonstrating our method’s strong detection and localization performance even when faced with complex novel anomalies significantly different from those encountered during training.

Strategies to Increase Financial Literacy for the People of Tirtawangunan Village

One of the problems faced by the community is financial management. Lack of knowledge about financial literacy can lead to various problems, such as difficulties in managing budgets, understanding investment concepts, managing debt, and planning for the financial future. To answer this, strategy training was carried out in improving financial literacy among the Tirtawangunan Village Community through community service activities with an assessment of success indicators and an evaluation of the results of the activity was carried out for 2 weeks after the activity took place. This is evident from the 25 trainees who have applied the use of financial record applications in managing finances more effectively and efficiently for those who are used as a strategy to improve financial literacy.

A Study on Filter-Based Adversarial Image Classification Models

In view of the fact that adversarial examples can lead to high-confidence erroneous outputs of deep neural networks, this study aims to improve the safety of deep neural networks by distinguishing adversarial examples. A classification model based on filter residual network structure is used to accurately classify adversarial examples. The filter-based classification model includes residual network feature extraction and classification modules, which are iteratively optimized by an adversarial training strategy. Three mainstream adversarial attack methods are improved, and adversarial samples are generated on the Mini-ImageNet dataset. Subsequently, these samples are used to attack the EfficientNet and the filter-based classification model respectively, and the attack effects are compared. Experimental results show that the filter-based classification model has high classification accuracy when dealing with Mini-ImageNet adversarial examples. Adversarial training can effectively enhance the robustness of deep neural network models.

Perceived Training Needs of the Informal Caregivers of Older Adults: A Cross-Sectional Study

Background: Due to the enormous caregiving burden faced by informal caregivers, providing appropriate skills training has become an important supporting strategy in many countries/regions. Understanding caregivers’ training needs is instrumental in designing effective training intervention programs, which are expected to reduce the caregiving burden of informal caregivers and avoid the health deterioration associated with caregiving. This paper aims to explore the potential training needs of informal caregivers in Shanghai, and to identify the factors associated with these perceived training needs. Methods: A total of 196 eligible informal caregivers participated in this survey. A multivariate analysis was conducted to explore the factors associated with informal caregivers’ perceived training needs. Results: 86.7% (N = 170) of the caregivers reported at least one need for targeted training activity, and 62.7% (N = 123) of them identified two or more training needs. The top three activities requiring training included the following: self-care skills; safety supervision; and functional rehabilitation. The factors associated with various training needs included the health status of the care recipient, complementary caregiving support, caregiving stress, and the personal attributes of the informal caregiver. The technical skills training needs were more related to the care recipients’ health status (e.g., dependency level, disease progression) and formal care support resources. Conversely, the intangible skills training needs were more sensitive to caregiver attributes (e.g., gender, age, and education level). Conclusions: A personalized training strategy and early-stage intervention program are critical to providing effective support to informal caregivers. The potential implications are to raise awareness of the importance of skills training for informal caregivers, and to inform the implementation of effective training strategies for improving the quality of informal care and the well-being of informal caregivers in China.

The impact of explicit and implicit instruction on EFL learners’ oral performance

Abstract This quasi-experimental study investigated the impacts of two instructional conditions on the fluency of learners’ performance. In the first condition (strategy training), learners were provided with explicit fluency strategy training for a portion of each lesson before rehearsing versions of the speaking task for the remainder of the lesson. In the second condition (massed practice), learners rehearsed the versions of the target speaking task for the entire lesson. Both instructional conditions allowed learners to benefit from language input and task repetition to help conceptualize and encode the content of their speech. Two intact classes of university EFL learners participated in this study. The data were audio recordings of participants’ oral performance of news report presentations in pre-test and post-test conditions. This data was transcribed and coded for speech rate, articulation rate, phonation/time ratio, number of filled and unfilled pauses, mean duration of pauses, mean length of run, verbatim repetition, and repairs. Although the overall effect of the training was found to be insignificant, the estimation analysis revealed that both instructional conditions improve participants’ speech fluency, with the effect sizes ranging from small to medium. These effect sizes are considered meaningful for the speech fluency development of EFL learners in the study context. Hence, both instructional conditions could be applied as a potentially complementary pedagogical strategy in EFL classrooms at the university level, particularly with regard to increasing the accuracy of teachers’ perceptions of learners’ oral proficiency.

Improving Model Robustness through Hybrid Adversarial Training: Integrating FGSM and PGD Methods

Abstract. With the widespread use of deep learning models in various applications. People are gradually realizing the vulnerability of these models to adversarial attacks. Adversarial training is an effective strategy to defend against adversarial attacks. Based on the advantages and disadvantages of the current mainstream Fast Gradient Sign Method (FGSM) adversarial training and Projected Gradient Descent (PGD) adversarial training, this paper proposes a hybrid adversarial training that integrates FGSM and PGD methods and uses the ResNet-18 model and SVHN dataset for testing. Experimental results show that hybrid adversarial training can effectively reduce training time. Its accuracy on the original data set is higher than that of PGD adversarial training, which is improved by about 2%. The performance when facing FGSM attacks is almost the same as that of single FGSM adversarial training. The performance when facing PGD attacks decreases more significantly, which is about 2% to 3% lower than that of PGD adversarial training. This study not only helps to understand the robustness of hybrid adversarial training to models facing adversarial attacks but also helps in studying new adversarial training strategies.

A Deep Learning Algorithm for KOL Segmentation on Social Media Videos

Nowadays, there is high commercial demand for product replacement which places the products virtually in Key Opinion Leader’s (KOL’s) social media videos. However, one of the challenges of placing the products virtually is the KOL segmentation. Since KOLs often hold products in front of them, it requires the segmentation to segment not only humans but also different products. This paper introduces the state-of-the-art deep learning method, namely RSUDISNet, for KOL segmentation. The proposed technique integrates two deep Convolutional Neural Network (CNN) technologies. One is the Matting Objective Decomposition Network (MODNet), which segments KOLs well but not the products blocking the KOLs. The other one is the two-level nested U-structure network (U2Net) based on the salient object detection method to segment the objects well, but not the KOL. The key technique of the proposed research is to employ the feature of the U2Net to embed the MODNet to overcome the problem of KOL segmentation. Since both MODNet and U2Net are lightweights, the combined network can be used for real-time scenarios. After that, the Intermediate Supervision (IS) training strategy is utilized to overcome the overfitting. The experimental results show that our proposed method outperforms the MODNet and U2Net.

Analyzing Targeted Muscle Strength: Impact on Speed, Endurance, and Performance in Female Volleyball

This study examines the influence of muscle strength on performance metrics in volleyball, addressing the need to understand how specific and non-specific strength training impacts athletic capabilities. A cohort of athletes underwent various strength tests, including squats (SQs), bench presses (BPs), and deadlifts (DLs), to assess their force and power during various forms of test. Lactate thresholds (LT1 and LT2) were also measured to evaluate aerobic and anaerobic capacities. The median speed at LT1 was 1.80 m/s, and LT2 was determined at a mean velocity of 2.56 m/s. Notably, a correlation was found between SQ performance and total points scored (p = 0.040, r = 0.285), while upper body strength was strongly correlated with performance metrics, showing significant relationships with points scored (p = 0.0001, r = 0.690) and peak power during BPs (p = 0.0001, r = 0.587). The findings suggest that targeted strength training enhances physical capabilities and improves technical and tactical performance in volleyball, highlighting the critical role of muscle strength in athlete selection and training strategies.

MP-GIEN: Vehicle Re-Identification Method Based on Multi-View Progressive Graph Interactive Embedding Network

Vehicle re-identification primarily faces two challenges: significant intra-class variations caused by different camera views and subtle inter-class differences between vehicles of the same model. In this paper, we propose a Multi-View Progressive Graph Interaction Embedding Network (MP-GIEN) for vehicle re-identification. First, we design a global semantic extractor. By employing Graph Interaction Units (GI Units) and semantic context, we combine semantic information to facilitate contextual reasoning of image regions, thereby extracting global semantic features of the target. The global loss is then calculated using ID loss and triplet loss. Second, we devise a local view-aware extractor. The U-Net network is utilized to parse the image into four views (front, rear, top, and side). Subsequently, features are aligned through mask average pooling, and a progressive training strategy is adopted to address the learning of fine-grained information. Lastly, we design a feature fusion enhancer. We revise the typical triplet loss to avoid mismatches in local features. By optimizing the local triplet loss and global loss, we learn the embedding of visual features, which not only reduces the intra-instance distance but also enlarges the inter-instance differences. MP-GIEN facilitates the capture of stable discriminative information on vehicles under different views. Experiments conducted on the VeRi-Wild dataset demonstrate that our model significantly outperforms state-of-the-art methods.