Performance Of Multiple Tasks Research Articles

A Multi-Source Domain Adaptation Method for Bearing Fault Diagnosis with Dynamically Similarity Guidance on Incomplete Data

In actual industrial scenarios, collecting a complete dataset with all fault categories under the same conditions is challenging, leading to a loss in fault category knowledge in single-source domains. Deep learning domain adaptation methods face difficulties in multi-source scenarios due to insufficient labeled data and significant distribution differences, hindering domain-specific knowledge transfer and reducing fault diagnosis efficiency. To address these issues, the Dynamic Similarity-guided Multi-source Domain Adaptation Network (DS-MDAN) is proposed. This method leverages incomplete data from multiple-source domains to address distribution disparities in deep domain adaptation. It enhances diagnostic performance in the target domain by transferring knowledge across diverse domains. DS-MDAN uses convolution kernels of different scales to extract multi-scale feature information and achieves feature fusion through upsampling and operations like addition and concatenation. Adversarial training with domain and fault classifiers optimizes feature extraction for widely applicable representations. The similarity between source and target domain data is calculated based on features extracted by a shared-weight network, dynamically adjusting the contribution of different source domain data to minimize distribution differences. Finally, matched source and target domain samples are mapped to the same feature space for fault diagnosis. Experimental validation on various bearing fault datasets shows that DS-MDAN improves performance in multiple fault diagnosis tasks, increasing accuracy and demonstrating good generalization capabilities.

Sleep Fragmentation Results in Novel Set-shifting Decrements in GCR-exposed Male and Female Rats.

The prolonged exposure to multiple spaceflight stressors during long-duration missions to the Moon and Mars will be challenging to the physical and mental health of the astronauts. Ground-based studies have reported that attentional set-shifting task (ATSET) performance is impaired after space radiation (SR) exposure. At certain times during deep-space missions, astronauts will likely have to contend with the combined impacts of SR and sleep perturbation. In rats, poor quality, fragmented sleep adversely impacts performance in multiple cognitive tasks, including the ATSET task. While both SR and sleep perturbations independently cause cognitive performance deficits, the incidence, severity and exact nature of those decrements following combined exposure to these flight stressors is largely unknown. This study established the impact that a single night of fragmented sleep has on ATSET performance in both male and female rats exposed to 10 cGy of galactic cosmic ray simulation (GCRsim). The GCRsim beam is a complex beam that mimics the mass and energy spectra of the SR particles that an astronaut will be exposed to within the spacecraft. Rats that had no obvious ATSET performance decrements when normally rested were subjected to fragmented sleep and their ATSET performance reassessed. Sleep fragmentation resulted in significant ATSET performance decrements in GCRsim-exposed rats, with specific performance decrements being observed in stages where attention or cue shifting is extensively used. Performance decrements in these stages are rarely observed after SR exposure. While both male and female rats exhibited latent sleep-related performance decrements, these were sex dependent, with male and female rats exhibiting different types of performance decrements (either reduced processing speed or task completion efficiency) in different stages of the ATSET task. This study suggests that SR-induced cognitive impairment may not be fully evident in normally rested rats, with an underestimation of both the incidence and nature of performance decrements that could occur when multiple space flight stressors are present. These data suggest that that there may be synergistic interactions between multiple space flight stressors that may not be easily predicted from their independent actions.

Application of CNN Classic Model in Modern Image Processing

As a deep learning model, convolutional neural network (CNN) has been widely used in the field of modern image processing and has shown excellent performance. From LeNet to AlexNet, to classic models such as VGGNet and ResNet, CNN has achieved remarkable success in tasks such as image classification, object detection and segmentation through multi-level feature extraction and automatic learning capabilities. This paper first explores the basic structure and working principle of CNN, analyzes the advantages and limitations of classic models, and reviews its specific applications in image processing. By introducing the optimization strategies of various models, it further explores the improvement path of CNN in the field of image processing, including model compression, lightweight design and the introduction of new network structures. In short, the continuous optimization of CNN has enabled it to show powerful performance in multiple complex tasks, promoted the rapid development of image processing technology, and provided strong support for applications in more fields.

Transferring knowledge by budget online learning for multiobjective multitasking optimization

Multiobjective multitasking optimization (MO-MTO) has attracted increasing attention in the evolutionary computation field. Evolutionary multitasking (EMT) algorithms can improve the overall performance of multiple multiobjective optimization tasks through transferring knowledge among tasks. Negative transfer resulting from the indeterminacy of the transferred knowledge may bring about the degradation of the algorithm performance. Identifying the valuable knowledge to transfer by learning the historical samples is a feasible way to reduce negative transfer. Taking this into account, this paper proposes a budget online learning based EMT algorithm for MO-MTO problems. Specifically, by regarding the historical transferred solutions as samples, a classifier would be trained to identified the valuable knowledge. The solutions which are considered containing valuable knowledge will have more opportunity to be transfer. For the samples arrive in the form of streaming data, the classifier would be updated in a budget online learning way during the evolution process to address the concept drift problem. Furthermore, the exceptional case that the classifier fails to identify the valuable knowledge is considered. Experimental results on two MO-MTO test suits show that the proposed algorithm achieves highly competitive performance compared with several traditional and state-of-the-art EMT methods.

Augmentation-aware self-supervised learning with conditioned projector

Self-supervised learning (SSL) is a powerful technique for learning from unlabeled data. By learning to remain invariant to applied data augmentations, methods such as SimCLR and MoCo can reach quality on par with supervised approaches. However, this invariance may be detrimental for solving downstream tasks that depend on traits affected by augmentations used during pretraining, such as color. In this paper, we propose to foster sensitivity to such characteristics in the representation space by modifying the projector network, a common component of self-supervised architectures. Specifically, we supplement the projector with information about augmentations applied to images. For the projector to take advantage of this auxiliary conditioning when solving the SSL task, the feature extractor learns to preserve the augmentation information in its representations. Our approach, coined Conditional Augmentation-aware Self-supervised Learning (CASSLE), is directly applicable to typical joint-embedding SSL methods regardless of their objective functions. Moreover, it does not require major changes in the network architecture or prior knowledge of downstream tasks. In addition to an analysis of sensitivity towards different data augmentations, we conduct a series of experiments, which show that CASSLE improves over various SSL methods, reaching state-of-the-art performance in multiple downstream tasks. 11We make our codebase available at https://github.com/gmum/CASSLE.22The supplementary material of this paper was shared separately due to page count limit (see Appendix A). The entire paper, along with the supplementary material, can also be accessed at https://arxiv.org/abs/2306.06082.33A short version of this paper appeared at the NeurIPS 2023 Workshop: Self-Supervised Learning-Theory and Practice (https://sslneurips23.github.io).

LGIT: local–global interaction transformer for low-light image denoising

Transformer-based methods effectively capture global dependencies in images, demonstrating outstanding performance in multiple visual tasks. However, existing Transformers cannot effectively denoise large noisy images captured under low-light conditions owing to (1) the global self-attention mechanism causing high computational complexity in the spatial dimension owing to a quadratic increase in computation with the number of tokens; (2) the channel-wise self-attention computation unable to optimise the spatial correlations in images. We propose a local–global interaction Transformer (LGIT) that employs an adaptive strategy to select relevant patches for global interaction, achieving low computational complexity in global self-attention computation. A top-N patch cross-attention model (TPCA) is designed based on superpixel segmentation guidance. TPCA selects top-N patches most similar to the target image patch and applies cross attention to aggregate information from them into the target patch, effectively enhancing the utilisation of the image’s nonlocal self-similarity. A mixed-scale dual-gated feedforward network (MDGFF) is introduced for the effective extraction of multiscale local correlations. TPCA and MDGFF were combined to construct a hierarchical encoder-decoder network, LGIT, to compute self-attention within and across patches at different scales. Extensive experiments using real-world image-denoising datasets demonstrated that LGIT outperformed state-of-the-art (SOTA) convolutional neural network (CNN) and Transformer-based methods in qualitative and quantitative results.

DBFiLM: A novel dual-branch frequency improved legendre memory forecasting model for coagulant dosage determination

In the process of drinking water preparation, the amount of coagulant added directly affects the quality of water and drinking safety. However, the coagulation process is influenced by many environmental factors, and the process of determining the dosage is complex and variable. The sedimentation reaction is severely delayed, making it difficult to establish a real-time coagulant dosage determination model. To address this issue, this study proposes a frequency-improved Legendre memory model based on deep learning. It has a dual-branch structure and accurately predicts the dosage of coagulant by learning the characteristic information in the historical time series. In the data instantiation module of the model, a local response mechanism is introduced to accurately describe the original distribution of the data by activating local information within the channel. In addition, the model adopts a dual-branch structure for parallel computation, capturing both local and global information of the sequence, and extracting richer information features from the historical data. With a lower computational cost, the model compensates for the loss of information in low-rank approximation operations, establishing longer time series mapping relationships. Experimental results verify that our proposed model has excellent performance, achieving a maximum improvement of 20.3% over the baseline Frequency improved Legendre Memory model in a real coagulant dosage dataset, and achieving optimal performance in multiple multi-step prediction tasks on a public dataset.

Many-to-many mapping: A simulation study of how the number of traits and tasks affect the evolution of form and function

Many-to-many mapping of form-to-function posits that multiple morphological and physiological traits affect the performance of multiple tasks in an organism, and that redundancy and multitasking occur simultaneously to shape the evolution of an organism’s phenotype. Many-to-many mapping is expected to be ubiquitous in nature, yet little is known about how it influences the evolution of organismal phenotype. The F-matrix is a powerful tool to study these issues because it describes how multiple traits affect multiple tasks. We undertook a simulation study using the F-matrix to test how the number of traits and the number of tasks affect trait integration and evolvability, as well as the relationships among tasks. We found that as the number of traits and/or tasks increases, the relationships between the tasks and the integration between the traits become weaker, and that the evolvability of the traits increases, all resulting in a system that is freer to evolve. We also found that as the number of traits increases, performance tradeoffs tend to become weaker, but only to a point. Our work shows that it is important to consider not only multiple traits, but also the multitude of tasks that those traits carry out when studying form-function relationships. We suggest that evolution of these relationships follows functional lines of least resistance, which are less defined in more complex systems, resulting in a mechanism for diversification.

SynCluster: Reaction Type Clustering and Recommendation Framework for Synthesis Planning.

AI-assisted synthesis planning has emerged as a valuable tool in accelerating synthetic chemistry for the discovery of new drugs and materials. The template-free approach, which showcases superior generalization capabilities, is seen as the mainstream direction in this field. However, it remains unclear whether such an end-to-end approach can achieve problem-solving performance on par with experienced chemists without fully revealing insights into the chemical mechanisms involved. Moreover, there is a lack of unified and chemically inspired frameworks for improving multitask reaction predictions in this area. In this study, we have addressed these challenges by investigating the impact of fine-grained reaction-type labels on multiple downstream tasks and propose a novel framework named SynCluster. This framework incorporates unsupervised clustering cues into the baseline models and identifies plausible chemical subspaces which is compatible with multitask extensions and can serve as model-independent indicators to effectively enhance the performance of multiple downstream tasks. In retrosynthesis prediction, SynCluster achieves significant improvements of 4.1 and 11.0% in top-1 and top-10 prediction accuracy, respectively, compared to the baseline Molecular Transformer, and achieves a notable enhancement of 13.9% in top-10 accuracy when combined with Retroformer. By incorporating simplified molecular-input line-entry system augmentation, our framework achieves higher top-10 accuracy compared to state-of-the-art sequence-based retrosynthesis models and improves over the baseline on the diversity and validity of reactants. SynCluster also achieves 94.9% top-10 accuracy in forward synthesis prediction and 51.5% top-10 Maxfrag accuracy in reagent prediction. Overall, SynCluster provides a fresh perspective with chemical interpretability and reinforcement of domain knowledge in the synthesis design. It offers a promising solution for improving the accuracy and efficiency of AI-assisted synthesis planning and bridges the gap between template-free approaches and the problem-solving abilities of experienced chemists.

Multimode microdimer robot for crossing tissue morphological barrier

Swimming microrobot energized by magnetic fields exhibits remotely propulsion and modulation in complex biological experiment with high precision. However, achieving high environment adaptability and multiple tasking capability in one configuration is still challenging. Here, we present a strategy that use oriented magnetized Janus spheres to assemble the microdimer robots with two magnetic distribution configurations of head-to-side configuration (HTS-config) and head-to-head configuration (HTH-config), achieving performance of multiple tasks through multimode transformation and locomotion. Modulating the magnetic frequency enables multimode motion transformation between tumbling, rolling, and swing motion with different velocities. The dual-asynchronization mechanisms of HTS-config and HTH-config robot dependent on magnetic dipole-dipole angle are investigated by molecular dynamic simulation. In addition, the microdimer robot can transport cell crossing morphological rugae or complete drug delivery on tissues by switching motion modes. This microdimer robot can provide versatile motion modes to address environmental variations or multitasking requirements.

Effectiveness of masked autoencoder in vision transformer models for image classification

This paper uses the literature reading method to systematically sort out and introduce the basic principles of the three algorithms of the transformer model and their application in the field of image classification, which has high theoretical value and social value, and has strong reference for the development of the transformer model in the future. ViT is simply an innovation in computer vision based on the transformer model. It first separates an image into several local patches (16x16), and then maps each one to a feature vector. These vectors will be delivered to an encoder for polishing. Finally, a special token is appended to these vectors for integrating location information. The final prediction is based on these tokesn. Swin-T is a new Transformer architecture, which is proposed by Microsoft Research to improve the performance of computer vision tasks. It adopts a new windowed feature extraction strategy, which can maintain high accuracy while significantly reducing the amount of computation and memory consumption. It has achieved leading performance in multiple computer vision tasks, becoming one of the most advanced visual Transformer models. In computer vision image classification, the information is highly redundant, the lack of an image piece, may not make the model produce much confusion, the model can be inferred from the surrounding pixel information, masked autoencoder (MAE) is to mask a high proportion of image pieces, create a difficult learning task, the method is simple but extremely effective.

Effects of MRI scanner manufacturers in classification tasks with deep learning models

Deep learning has become a leading subset of machine learning and has been successfully employed in diverse areas, ranging from natural language processing to medical image analysis. In medical imaging, researchers have progressively turned towards multi-center neuroimaging studies to address complex questions in neuroscience, leveraging larger sample sizes and aiming to enhance the accuracy of deep learning models. However, variations in image pixel/voxel characteristics can arise between centers due to factors including differences in magnetic resonance imaging scanners. Such variations create challenges, particularly inconsistent performance in machine learning-based approaches, often referred to as domain shift, where the trained models fail to achieve satisfactory or improved results when confronted with dissimilar test data. This study analyzes the performance of multiple disease classification tasks using multi-center MRI data obtained from three widely used scanner manufacturers (GE, Philips, and Siemens) across several deep learning-based networks. Furthermore, we investigate the efficacy of mitigating scanner vendor effects using ComBat-based harmonization techniques when applied to multi-center datasets of 3D structural MR images. Our experimental results reveal a substantial decline in classification performance when models trained on one type of scanner manufacturer are tested with data from different manufacturers. Moreover, despite applying ComBat-based harmonization, the harmonized images do not demonstrate any noticeable performance enhancement for disease classification tasks.

Synaptotagmin-1-dependent phasic axonal dopamine release is dispensable for basic motor behaviors in mice

In Parkinson’s disease (PD), motor dysfunctions only become apparent after extensive loss of DA innervation. This resilience has been hypothesized to be due to the ability of many motor behaviors to be sustained through a diffuse basal tone of DA; but experimental evidence for this is limited. Here we show that conditional deletion of the calcium sensor synaptotagmin-1 (Syt1) in DA neurons (Syt1 cKODA mice) abrogates most activity-dependent axonal DA release in the striatum and mesencephalon, leaving somatodendritic (STD) DA release intact. Strikingly, Syt1 cKODA mice showed intact performance in multiple unconditioned DA-dependent motor tasks and even in a task evaluating conditioned motivation for food. Considering that basal extracellular DA levels in the striatum were unchanged, our findings suggest that activity-dependent DA release is dispensable for such tasks and that they can be sustained by a basal tone of extracellular DA. Taken together, our findings reveal the striking resilience of DA-dependent motor functions in the context of a near-abolition of phasic DA release, shedding new light on why extensive loss of DA innervation is required to reveal motor dysfunctions in PD.

Toward Robust Diagnosis: A Contour Attention Preserving Adversarial Defense for COVID-19 Detection

As the COVID-19 pandemic puts pressure on healthcare systems worldwide, the computed tomography image based AI diagnostic system has become a sustainable solution for early diagnosis. However, the model-wise vulnerability under adversarial perturbation hinders its deployment in practical situation. The existing adversarial training strategies are difficult to generalized into medical imaging field challenged by complex medical texture features. To overcome this challenge, we propose a Contour Attention Preserving (CAP) method based on lung cavity edge extraction. The contour prior features are injected to attention layer via a parameter regularization and we optimize the robust empirical risk with hybrid distance metric. We then introduce a new cross-nation CT scan dataset to evaluate the generalization capability of the adversarial robustness under distribution shift. Experimental results indicate that the proposed method achieves state-of-the-art performance in multiple adversarial defense and generalization tasks. The code and dataset are available at https://github.com/Quinn777/CAP.

ChatGPT - još jedan korak bliže digitalnoj eri ili pretnja osnovnim pravima i slobodama?

Artificial Intelligence (AI) constitutes one of the most fundamental pillars for the implementation of the EU Digital Agenda. It corresponds to the tremendous ongoing technological evolution which is marked by the spread of the digitalization in both private and public sector. AI tools provide numerous services, such as faster decision-making, performance of multiple tasks and repetitive jobs on our behalf and diagnosis of risky situations. This paper puts a special emphasis on the ChatGPT which is considered the most illustrative representative of the current AI technology. Within a minimal time of its existence this innovative viral chatbot has started to dominate the world of AI. However, its use raises serious legal and ethical risks for our privacy and protection of fundamental rights and freedoms, born by the lack of a binding regulatory framework governing AI. Therefore, at first level, this study focuses on the legal regime which governs the use of ChatGPT, by interpreting the legal status, after giving a short demonstration of its function and services (Section I). Secondly, a critical approach will be pursued focusing on special issues regarding this new AI tool on the basis of its application in practice at the area of journalism (Section II). Following that intense analysis, the paper aims to lead to fruitful and original conclusions with the ultimate goal to enhance the establishment of a powerful, safe and trustful digital environment.

Self-Supervised Pretraining Improves Performance and Inference Efficiency in Multiple Lung Ultrasound Interpretation Tasks

Self-Supervised Pretraining Improves Performance and Inference Efficiency in Multiple Lung Ultrasound Interpretation Tasks

Uncertainty-Aware Clustering for Unsupervised Domain Adaptive Object Re-Identification

Unsupervised Domain Adaptive (UDA) object re-identification (Re-ID) aims at adapting a model trained on a labeled source domain to an unlabeled target domain. State-of-the-art object Re-ID approaches adopt clustering algorithms to generate pseudo-labels for the unlabeled target domain. However, the inevitable label noise caused by the clustering procedure significantly degrades the discriminative power of Re-ID model. To address this problem, we propose an uncertainty-aware clustering framework (UCF) for UDA tasks. First, a novel hierarchical clustering scheme is proposed to promote clustering quality. Second, an uncertainty-aware collaborative instance selection method is introduced to select images with reliable labels for model training. Combining both techniques effectively reduces the impact of noisy labels. In addition, we introduce a strong baseline that features a compact contrastive loss. Our UCF method consistently achieves state-of-the-art performance in multiple UDA tasks for object Re-ID, and significantly reduces the performance gap between unsupervised and supervised Re-ID. In particular, the performance of our unsupervised UCF method in the MSMT17->Market1501 task is better than that of the fully supervised setting on Market1501. The code of UCF is available at https://github.com/Wang-pengfei/UCF.

Ergonomic evaluation of the chicken hanger workplace in a bird processing company

Introduction: in the poultry processing industry, activities are developed that require the performance of multiple repetitive tasks, generally while standing upright; which increases the risk of developing musculoskeletal disorders in the upper extremities and lower back. These risk factors include: repetitive movements, forced postures, awkward and static postures, vibration and high and low temperatures. Affecting body segments such as neck, shoulders and back. Aim: To ergonomically evaluate the work station: "chicken hanger" of a poultry processing company, in the state of Aragua-Venezuela, during the year 2022. Methods: Quantitative, descriptive, non-experimental, field and cross-sectional research. The population and census sample is constituted by eight (8) workers. In the collection of information, the interview was applied to obtain sociodemographic and labor data, in the description of the work position the documented observation technique was used and for the determination of musculoskeletal discomfort the body schema of pain and the Check List OCRA method were used. Result: the entire population is male, right-handed, average age 33 ±2.67, normal weight, average seniority 4 years ±2.36, with pain at the end of the workday in the shoulder, lower back, upper back and knees. Their Ocra Index was high on the right side and medium on the left side. Conclusion: in the job "chicken hanger" there is a high risk of suffering musculoskeletal disorders in the upper limbs, with greater relevance in the right upper limb

ScSemiGAN: a single-cell semi-supervised annotation and dimensionality reduction framework based on generative adversarial network.

Cell-type annotation plays a crucial role in single-cell RNA-seq (scRNA-seq) data analysis. As more and more well-annotated scRNA-seq reference data are publicly available, automatical label transference algorithms are gaining popularity over manual marker gene-based annotation methods. However, most existing methods fail to unify cell-type annotation with dimensionality reduction and are unable to generate deep latent representation from the perspective of data generation. In this article, we propose scSemiGAN, a single-cell semi-supervised cell-type annotation and dimensionality reduction framework based on a generative adversarial network, to overcome these challenges, modeling scRNA-seq data from the aspect of data generation. Our proposed scSemiGAN is capable of performing deep latent representation learning and cell-type label prediction simultaneously. Through extensive comparison with four state-of-the-art annotation methods on diverse simulated and real scRNA-seq datasets, scSemiGAN achieves competitive or superior performance in multiple downstream tasks including cell-type annotation, latent representation visualization, confounding factor removal and enrichment analysis. The code and data of scSemiGAN are available on GitHub: https://github.com/rafa-nadal/scSemiGAN. Supplementary data are available at Bioinformatics online.

Explicit Metric-Based Multiconcept Multi-Instance Learning With Triplet and Superbag.

Multi-instance learning (MIL) has garnered considerable attention in recent years due to its favorable performance in various scenarios. Nonetheless, most previous studies have implicitly expressed the correlation between instances and bags. Moreover, the importance of negative instances has been largely overlooked. Hence, we seek to present an explicit and intuitively understandable method that can compensate for these deficiencies. In this article, we creatively introduce a metric-based multiconcept MIL approach based on two aspects. First, the triplet-based bag embedding method identifies instance categories and builds attention weights for every instance explicitly. Accordingly, bag embedding is accomplished under the limitation of weak supervision. Second, the developed instance correlation metric approach in the superbag considers the multiconcept issue to boost the model generalization performance. We have designed a rich variety of experiments to demonstrate the performance of our algorithm. The artificial data experiment reveals the interpretability of the proposed network. The results of the comparison experiment confirm that our method shows favorable performance in multiple tasks. Finally, we illustrate the motivation of the presented method by the ablation experiments.