Noisy Datasets Research Articles

CAS3D: 3D quantitative morphometry on Second Harmonic Generation image volumes from single skeletal muscle fibers

The CAS3D image processing method intuitively applies a combination of Fourier space and real space 3D analysis algorithms to volumetric images of single skeletal muscle fiber Myosin II Second Harmonic Generation (SHG) XYZ image data. Our developed tool automatically quantifies the myofibrillar orientation in muscle samples by determining the cosine angle sum of intensity gradients in 3D (CAS3D) while determining the mean sarcomere length (SL) and sample orientation. The expected CAS3D values could be reproduced from ideal artificial data sets. Applied random noise in artificial images lowers the detected CAS3D value, and for noise levels below 20%, the correlation can be approximated by a linear function with a slope of −0.006 CAS3D/noise%. The deviations in SL and orientation detection were determined on ideal and noisy artificial data sets and were statistically indistinguishable from 0 (null hypothesis t-test P > 0.1). The software was applied to a previously published data set of single skeletal muscle fiber volumetric SHG image data from a rat intensive care unit (ICU) model of ventilator-induced diaphragm dysfunction (VIDD) with treatment regimens involving the small anti-inflammatory molecules BGP-15, vamorolone, or prednisolone. Our method reliably reproduced the results of the previous work and improved the standard deviation of the cosine angle sum detection in all sample groups from a mean of 0.03 to 0.008. This improvement is achieved by applying analysis algorithms to the whole volumetric images in 3D in contrast to the previously common method of slice-wise XY analysis.

An Adaptive Xception Model for Classification of Brain Tumors

Classification of different brain tumors is challenging due to unpredictable variations in intra-inter-classes. Unlike existing methods which are not effective for images of complex backgrounds, the proposed work aims at accurate classification of diverse types of brain tumors such that an appropriate model can be used for disease identification. This study considers glioma, meningioma, no tumor, and pituitary tumors for classification. To achieve an accurate classification, we explore the Xception architecture layer, which involves flattening, dropout, and dense layer operations. The model extracts features based on shapes, spatial relationships, and structure of the image, discriminating between the different brain tumor images. The model is evaluated on a dataset of 7023 MRI images for classification. The results of a large dataset and comparative study with the existing methods show that the proposed method is better than state of the art in terms of classification rate. Specifically, our method achieves more than a 90% average classification rate, which is better than state of the art. The results on noisy and blurred datasets show that the proposed model is robust to noise and blur.

Comparing human text classification performance and explainability with large language and machine learning models using eye-tracking

To understand the alignment between reasonings of humans and artificial intelligence (AI) models, this empirical study compared the human text classification performance and explainability with a traditional machine learning (ML) model and large language model (LLM). A domain-specific noisy textual dataset of 204 injury narratives had to be classified into 6 cause-of-injury codes. The narratives varied in terms of complexity and ease of categorization based on the distinctive nature of cause-of-injury code. The user study involved 51 participants whose eye-tracking data was recorded while they performed the text classification task. While the ML model was trained on 120,000 pre-labelled injury narratives, LLM and humans did not receive any specialized training. The explainability of different approaches was compared based on the top words they used for making classification decision. These words were identified using eye-tracking for humans, explainable AI approach LIME for ML model, and prompts for LLM. The classification performance of ML model was observed to be relatively better than zero-shot LLM and non-expert humans, overall, and particularly for narratives with high complexity and difficult categorization. The top-3 predictive words used by ML and LLM for classification agreed with humans to a greater extent as compared to later predictive words.

A novel method combining deep learning with the Kennard-Stone algorithm for training dataset selection for image-based rice seed variety identification.

Different varieties of rice vary in planting time, stress resistance, and other characteristics. With advances in rice-breeding technology, the number of rice varieties has increased significantly, making variety identification crucial for both trading and planting. This study collected RGB images of 20 hybrid rice seed varieties. An enhanced deep super-resolution network (EDSR) was employed to enhance image resolution, and a variety classification model utilizing the high-resolution dataset demonstrated superior performance to that of the model using the low-resolution dataset. A novel training sample selection methodology was introduced integrating deep learning with the Kennard-Stone (KS) algorithm. Convolutional neural networks (CNN) and autoencoders served as supervised and unsupervised feature extractors, respectively. The extracted feature vectors were subsequently processed by the KS algorithm to select training samples. The proposed methodologies exhibited superior performance over the random selection approach in rice variety classification, with an approximately 10.08% improvement in overall classification accuracy. Furthermore, the impact of noise on the proposed methodology was investigated by introducing noise to the images, and the proposed methodologies maintained superior performance relative to the random selection approach on the noisy image dataset. The experimental results indicate that both supervised and unsupervised learning models performed effectively as feature extractors, and the deep learning framework significantly influenced the selection of training set samples. This study presents a novel approach for training sample selection in classification tasks and suggests the potential for extending the proposed method to image datasets and other types of datasets. Further exploration of this potential is warranted. © 2024 Society of Chemical Industry.

RobU-Net: a heuristic robust multi-class brain tumor segmentation approaches for MRI scans

A tumor is an abnormal growth of cells, either cancerous or benign, that develops in an organ. Early detection and segmentation of brain tumors are crucial for effective treatment, but manual analysis by experts is a labor-intensive and time-consuming process. The proposed solution is a deep learning framework called RobU-Net, a modified U-Net, to improve the accuracy and robustness of MRI image segmentation, specifically designed to handle Rician noise in MRI scans. The study uses various encoder-decoder architectures to segment MRI scans, achieving optimal results that match the ground truth. The approach also employs discrete wavelet transforms for contrast enhancement and tests the framework's robustness. This study used the MRI brain tumor dataset with 3064 slices. The space and time analysis have been carried out using sensitivity (Sn), dice loss (DL), dice coefficient (DC), and Jaccard index (JI). A comprehensive assessment of our proposed framework depicted that it is better than recently existing frameworks. We have found the so far highest Sn, DC, and JI values for noise-impregnated and original MRI datasets using two novice architectures (overall best for A1-E2 and optimum or noisy dataset for A2-E1) as (0.9831, 0.9726, and 0.9468) and (0.9199, 0.9781, and 0.9571).

3WAUS: A novel three-way adaptive uncertainty-suppressing model for facial expression recognition

Facial Expression Recognition (FER) remains a challenging task due to the uncertainty caused by the noisy labels, interrater disagreement, and the inter-class similarity of facial expressions. However, the majority of existing FER methods are deterministic, resulting in limited generalization performance in dealing with uncertainty. By adding a non-deterministic option, three-way decision (3WD) enhances the ability to handle uncertainty, which is widely used in different domains to reduce decision risk. Therefore, we introduce 3WD model into FER, and propose a novel three-way adaptive uncertainty-suppressing model (3WAUS). The model divides the sample space into three regions: high-importance, medium-importance, and low-importance, with different action strategies to effectively suppress the uncertainty in FER. Specifically, 3WAUS utilizes a weighted learning strategy to reinforce learning in the high-importance (low-uncertainty) region, thus acquiring high-quality sample features. For medium-importance samples, a dynamic adaptive re-labeling strategy is proposed to modify noisy labels, enabling the model to learn more accurate representations and avoid overfitting noise. Additionally, 3WAUS also attenuates the learning for low-importance (high-uncertainty) samples, effectively reducing the misguidance of noise on the model. Finally, extensive experiments on three public datasets and synthetic noisy datasets indicate that the proposed 3WAUS outperforms state-of-the-art methods in dealing with uncertainty.

Revisiting Bundle Recommendation for Intent-aware Product Bundling

Product bundling represents a prevalent marketing strategy in both offline stores and e-commerce systems. Despite its widespread use, previous studies on bundle recommendation face two significant limitations. Firstly, they rely on noisy datasets, where bundles are defined by heuristics, e.g., products co-purchased in the same session. Secondly, they target specific tasks by holding unrealistic assumptions, e.g., the availability of bundles for recommendation directly. This paper proposes to take a step back and considers the process of bundle recommendation from a holistic user experience perspective. We first construct high-quality bundle datasets with rich metadata, particularly bundle intents, through a carefully designed crowd-sourcing task. We then define a series of tasks that together, support all key steps in a typical bundle recommendation process, from bundle detection, completion and ranking, to explanation and auto-naming, whereby 19 research questions are raised correspondingly to guide the analysis. Finally, we conduct extensive experiments and analyses with representative recommendation models and large language models (LLMs), demonstrating the challenges and opportunities, especially with the emergence of LLMs. To summarize, our study contributes by introducing novel data sources, paving the way for new research avenues, and offering insights to guide product bundling in real e-commerce platforms.

Test Input Prioritization for 3D Point Clouds

3D point cloud applications have become increasingly prevalent in diverse domains, showcasing their efficacy in various software systems. However, testing such applications presents unique challenges due to the high-dimensional nature of 3D point cloud data and the vast number of possible test cases. Test input prioritization has emerged as a promising approach to enhance testing efficiency by prioritizing potentially misclassified test cases during the early stages of the testing process. Consequently, this enables the early labeling of critical inputs, leading to a reduction in the overall labeling cost. However, applying existing prioritization methods to 3D point cloud data is constrained by several factors: (1) inadequate consideration of crucial spatial information, and (2) susceptibility to noises inherent in 3D point cloud data. In this article, we propose PCPrior , the first test prioritization approach specifically designed for 3D point cloud test cases. The fundamental concept behind PCPrior is that test inputs closer to the decision boundary of the model are more likely to be predicted incorrectly. To capture the spatial relationship between a point cloud test and the decision boundary, we propose transforming each test (a point cloud) into a low-dimensional feature vector, toward indirectly revealing the underlying proximity between a test and the decision boundary. To achieve this, we carefully design a group of feature generation strategies, and for each test input, we generate four distinct types of features, namely spatial features, mutation features, prediction features, and uncertainty features. Through a concatenation of the four feature types, PCPrior assembles a final feature vector for each test. Subsequently, a ranking model is employed to estimate the probability of misclassification for each test based on its feature vector. Finally, PCPrior ranks all tests based on their misclassification probabilities. We conducted an extensive study based on 165 subjects to evaluate the performance of PCPrior, encompassing both natural and noisy datasets. The results demonstrate that PCPrior outperforms all of the compared test prioritization approaches, with an average improvement of 10.99% to 66.94% on natural datasets and 16.62% to 53% on noisy datasets.

Submarine cable positioning using a residual convolutional neural network based on magnetic features

Accurate positioning is important for improving the efficiency of repairing submarine cables and reducing the related repair costs. The magnetic anomaly produced by a submarine cable can be used to estimate its vertical and horizontal positions. A novel approach using magnetic data for estimating the position of submarine cables based on the 1D residual convolutional neural network (RCNN) is investigated. Infinitely long ferromagnetic cylinder models with different parameters are used to generate data sets for model training and testing. Tests on noisy synthetic data sets show that the developed 1D RCNN method can capture detailed features related to the magnetic source position information, which is more accurate than the conventional Euler method in estimating the position of submarine cables. The developed 1D RCNN method has also been successfully applied to processing field data. Furthermore, the processing workflow of our 1D RCNN method is less noise-sensitive compared with the conventional Euler method. The proposed 1D RCNN method and its workflow open a new window for estimating the position of submarine cables using magnetic data.

Geometry effects in Impact Echo Signals - why traditional time-frequency filtering fails to remove them

Impact Echo signals can sometimes contain a lot of noise, even though the wavelength of the investigated modes is in the order of the specimen thickness. A special type of noise signals are the so-called geometry effects. These effects occur when the lateral dimensions of the investigated structure are not much larger than its thickness. This study compares signals with and without geometry effects using various types of signal processing methods such as classical fast Fourier transform, short-time Fourier transform, and the continuous wavelet transform. Results show that signals without geometry effects have a much more stable resonance frequency, whereas the main frequency content in Impact Echo signals containing geometry effects typically oscillates between two or more discrete frequencies. As this oscillation lasts for the entirety of the signals, traditional time-frequency filters will likely not be capable to remove geometry effects from Impact Echo signals. Resonance frequencies in Impact Echo signals unaffected by geometry effects are much more stable, with no fluctuation or oscillation over time. By comparing time-frequency representations of signals obtained at the same spot with different impactors, it is observed that the peak frequency at the beginning of each signal is similar for all three measurements. This result suggests that, for signals heavily affected by geometry effects, the beginning of the signal, which is usually neglected during Impact Echo inspection, could provide useful information regarding the thickness of the inspected structure. Two ways of obtaining thickness information from the signal beginning are presented and results are compared using a dataset from a reference specimen: one method uses classical Fourier transform of the signal beginning and the other approach uses the maximum frequency determined from the continuous wavelet transform. The results for both approaches show that the Impact Echo signal beginning contains depth information, which could prove valuable for evaluations of noisy datasets.

Speech Emotion Recognition

Speech emotion recognition is a vital area of research with applications starting from human-computer interaction to mental health monitoring. This paper will provide a comprehensive survey of the techniques, methods, applications, and challenges in speech emotion recognition. It begins by checking the significance of recognizing emotions from speech and its diverse applications across various fields. The paper then shows the method employed for emotion speech recognition, encompassing traditional machine learning techniques, such as support vector machines and Gaussian mixture models, as well as contemporary approaches, including deep learning and multimodal fusion. Moreover, it examines benchmark datasets commonly used for training and evaluation purposes in emotion speech recognition research. Speech Emotion Recognition (SER) has a wide range of applications and there has been a lot of research going on in this fascinating area in recent years. However, the entertainment sector suffers from a lack of study in this research. Many of use Neural Network (NN) and Long Short-Term Memory (LSTM) architectures to categorize the emotions in audio recordings captured by actors expressing various emotions. Moreover, our survey explores real-world applications of emotion speech recognition like virtual assistants, health sector, market sector, education, Mental health diagnosis. The paper discusses the challenge associated with emotion speech recognition, including the variability of emotional expressions, cultural influences, humor and privacy concerns. In future it will help others to deal with many Noisy dataset and various cultural effects of emotion.

Multitask Learning Strategy with Pseudo-Labeling: Face Recognition, Facial Landmark Detection, and Head Pose Estimation

Most facial analysis methods perform well in standardized testing but not in real-world testing. The main reason is that training models cannot easily learn various human features and background noise, especially for facial landmark detection and head pose estimation tasks with limited and noisy training datasets. To alleviate the gap between standardized and real-world testing, we propose a pseudo-labeling technique using a face recognition dataset consisting of various people and background noise. The use of our pseudo-labeled training dataset can help to overcome the lack of diversity among the people in the dataset. Our integrated framework is constructed using complementary multitask learning methods to extract robust features for each task. Furthermore, introducing pseudo-labeling and multitask learning improves the face recognition performance by enabling the learning of pose-invariant features. Our method achieves state-of-the-art (SOTA) or near-SOTA performance on the AFLW2000-3D and BIWI datasets for facial landmark detection and head pose estimation, with competitive face verification performance on the IJB-C test dataset for face recognition. We demonstrate this through a novel testing methodology that categorizes cases as soft, medium, and hard based on the pose values of IJB-C. The proposed method achieves stable performance even when the dataset lacks diverse face identifications.

Fast computational approach with prior dimension reduction for three-dimensional chemical component analysis using CT data of spectral imaging.

Spectral image (SI) measurement techniques, such as X-ray absorption fine structure (XAFS) imaging and scanning transmission electron microscopy (STEM) with energy-dispersive X-ray spectroscopy (EDS) or electron energy loss spectroscopy (EELS), are useful for identifying chemical structures in composite materials. Machine-learning techniques have been developed for automatic analysis of SI data, and their usefulness has been proven. Recently, an extended measurement technique combining SI with a computed tomography (CT) technique (CT-SI), such as CT-XAFS and STEM-EDS/EELS tomography, was developed to identify the three-dimensional (3D) structures of chemical components. CT-SI analysis can be conducted by combining CT reconstruction algorithms and chemical component analysis based on machine learning techniques. However, this analysis incurs high computational costs owing to the size of the CT-SI datasets. To address this problem, this study proposed a fast computational approach for 3D chemical component analysis in an unsupervised learning setting. The primary idea for reducing the computational cost involved compressing the CT-SI data prior to CT computation and performing 3D reconstruction and chemical component analysis on the compressed data. The proposed approach significantly reduced the computational cost without losing information about the 3D structure and chemical components. We experimentally evaluated the proposed approach using synthetic and real CT-XAFS data, which demonstrated that our approach achieved a significantly faster computational speed than the conventional approach while maintaining analysis performance. As the proposed procedure can be implemented with any CT algorithm, it is expected to accelerate 3D analyses with sparse regularized CT algorithms in noisy and sparse CT-SI datasets.

A Direct Approach for Local Quasi-Geoid Modeling Based on Spherical Radial Basis Functions Using a Noisy Satellite-Only Global Gravity Field Model

A Direct Approach for Local Quasi-Geoid Modeling Based on Spherical Radial Basis Functions Using a Noisy Satellite-Only Global Gravity Field Model

Suppressing label noise in medical image classification using mixup attention and self-supervised learning

Deep neural networks (DNNs) have been widely applied in medical image classification and achieve remarkable classification performance. These achievements heavily depend on large-scale accurately annotated training data. However, label noise is inevitably introduced in the medical image annotation, as the labeling process heavily relies on the expertise and experience of annotators. Meanwhile, DNNs suffer from overfitting noisy labels, degrading the performance of models. Therefore, in this work, we innovatively devise a noise-robust training approach to mitigate the adverse effects of noisy labels in medical image classification. Specifically, we incorporate contrastive learning and intra-group mixup attention strategies into vanilla supervised learning. The contrastive learning for feature extractor helps to enhance visual representation of DNNs. The intra-group mixup attention module constructs groups and assigns self-attention weights for group-wise samples, and subsequently interpolates massive noisy-suppressed samples through weighted mixup operation. We conduct comparative experiments on both synthetic and real-world noisy medical datasets under various noise levels. Rigorous experiments validate that our noise-robust method with contrastive learning and mixup attention can effectively handle with label noise, and is superior to state-of-the-art methods. An ablation study also shows that both components contribute to boost model performance. The proposed method demonstrates its capability of curb label noise and has certain potential toward real-world clinic applications.

Estimating a Function and Its Derivatives Under a Smoothness Condition

We consider the problem of estimating an unknown function [Formula: see text] and its partial derivatives from a noisy data set of n observations, where we make no assumptions about [Formula: see text] except that it is smooth in the sense that it has square integrable partial derivatives of order m. A natural candidate for the estimator of [Formula: see text] in such a case is the best fit to the data set that satisfies a certain smoothness condition. This estimator can be seen as a least squares estimator subject to an upper bound on some measure of smoothness. Another useful estimator is the one that minimizes the degree of smoothness subject to an upper bound on the average of squared errors. We prove that these two estimators are computable as solutions to quadratic programs, establish the consistency of these estimators and their partial derivatives, and study the convergence rate as [Formula: see text]. The effectiveness of the estimators is illustrated numerically in a setting where the value of a stock option and its second derivative are estimated as functions of the underlying stock price.

Spatial domain decomposition-based physics-informed neural networks for practical acoustic propagation estimation under ocean dynamics.

Practical acoustic propagation modeling is significantly affected by ocean dynamics, and then can be exploited in numerous oceanic applications, where "practical" refers to modeling acoustic propagation in simulations that mimic real-world ocean environments. Physics-based numerical models provide approximate solutions of wave equation and rely on accurate prior environmental knowledge while the environment of practical scenarios is largely unknown. In contrast, data-driven machine learning offers a promising avenue to estimate practical acoustic propagation by learning from dataset. However, collecting such a high-quality, noise-free, and dense dataset remains challenging. Under the practical hurdle posed by the above approaches, the emergence of physics-informed neural network (PINN) presents an alternative to integrate physics and data but with limited representation capacity. In this work, a framework, termed spatial domain decomposition-based physics-informed neural networks (SPINNs), is proposed to enhance the representation capacity in spatially dependent oceanic scenarios and effectively learn from incomplete and biased prior physics and noisy dataset. Experiments demonstrate SPINNs' advantages over PINN in practical acoustic propagation estimation. The learning capacity of SPINNs toward physics and dataset during training is further analyzed. This work holds promise for practical applications and future expansion.

Knockoffs-SPR: Clean Sample Selection in Learning With Noisy Labels.

A noisy training set usually leads to the degradation of the generalization and robustness of neural networks. In this article, we propose a novel theoretically guaranteed clean sample selection framework for learning with noisy labels. Specifically, we first present a Scalable Penalized Regression (SPR) method, to model the linear relation between network features and one-hot labels. In SPR, the clean data are identified by the zero mean-shift parameters solved in the regression model. We theoretically show that SPR can recover clean data under some conditions. Under general scenarios, the conditions may be no longer satisfied; and some noisy data are falsely selected as clean data. To solve this problem, we propose a data-adaptive method for Scalable Penalized Regression with Knockoff filters (Knockoffs-SPR), which is provable to control the False-Selection-Rate (FSR) in the selected clean data. To improve the efficiency, we further present a split algorithm that divides the whole training set into small pieces that can be solved in parallel to make the framework scalable to large datasets. While Knockoffs-SPR can be regarded as a sample selection module for a standard supervised training pipeline, we further combine it with a semi-supervised algorithm to exploit the support of noisy data as unlabeled data. Experimental results on several benchmark datasets and real-world noisy datasets show the effectiveness of our framework and validate the theoretical results of Knockoffs-SPR.

Robust multi-frequency band joint dictionary learning with low-rank representation

Emotional state recognition is an important part of emotional research. Compared to non-physiological signals, the electroencephalogram (EEG) signals can truly and objectively reflect a person’s emotional state. To explore the multi-frequency band emotional information and address the noise problem of EEG signals, this paper proposes a robust multi-frequency band joint dictionary learning with low-rank representation (RMBDLL). Based on the dictionary learning, the technologies of sparse and low-rank representation are jointly integrated to reveal the intrinsic connections and discriminative information of EEG multi-frequency band. RMBDLL consists of robust dictionary learning and intra-class/inter-class local constraint learning. In robust dictionary learning part, RMBDLL separates complex noise in EEG signals and establishes clean sub-dictionaries on each frequency band to improve the robustness of the model. In this case, different frequency data obtains the same encoding coefficients according to the consistency of emotional state recognition. In intra-class/inter-class local constraint learning part, RMBDLL introduces a regularization term composed of intra-class and inter-class local constraints, which are constructed from the local structural information of dictionary atoms, resulting in intra-class similarity and inter-class difference of EEG multi-frequency bands. The effectiveness of RMBDLL is verified on the SEED dataset with different noises. The experimental results show that the RMBDLL algorithm can maintain the discriminative local structure in the training samples and achieve good recognition performance on noisy EEG emotion datasets.

Cluster-Based Pairwise Contrastive Loss for Noise-Robust Speech Recognition.

This paper addresses a joint training approach applied to a pipeline comprising speech enhancement (SE) and automatic speech recognition (ASR) models, where an acoustic tokenizer is included in the pipeline to leverage the linguistic information from the ASR model to the SE model. The acoustic tokenizer takes the outputs of the ASR encoder and provides a pseudo-label through K-means clustering. To transfer the linguistic information, represented by pseudo-labels, from the acoustic tokenizer to the SE model, a cluster-based pairwise contrastive (CBPC) loss function is proposed, which is a self-supervised contrastive loss function, and combined with an information noise contrastive estimation (infoNCE) loss function. This combined loss function prevents the SE model from overfitting to outlier samples and represents the pronunciation variability in samples with the same pseudo-label. The effectiveness of the proposed CBPC loss function is evaluated on a noisy LibriSpeech dataset by measuring both the speech quality scores and the word error rate (WER). The experimental results reveal that the proposed joint training approach using the described CBPC loss function achieves a lower WER than the conventional joint training approaches. In addition, it is demonstrated that the speech quality scores of the SE model trained using the proposed training approach are higher than those of the standalone-SE model and SE models trained using conventional joint training approaches. An ablation study is also conducted to investigate the effects of different combinations of loss functions on the speech quality scores and WER. Here, it is revealed that the proposed CBPC loss function combined with infoNCE contributes to a reduced WER and an increase in most of the speech quality scores.