Data Augmentation Algorithms Research Articles

Spatial Variation Generation Algorithm for Motor Imagery Data Augmentation: Increasing the Density of Sample Vicinity.

The imbalanced development between deep learning-based model design and motor imagery (MI) data acquisition raises concerns about the potential overfitting issue- models can identify training data well but fail to generalize test data. In this study, a Spatial Variation Generation (SVG) algorithm for MI data augmentation is proposed to alleviate the overfitting issue. In essence, SVG generates MI data using variations of electrode placement and brain spatial pattern, ultimately elevating the density of the raw sample vicinity. The proposed SVG prevents models from memorizing the training data by replacing the raw samples with the proper vicinal distribution. Moreover, SVG generates a uniform distribution and stabilizes the training process of models. In comparison studies involving five deep learning-based models across eight datasets, the proposed SVG algorithm exhibited a notable improvement of 0.021 in the area under the receiver operating characteristic curve (AUC). The improvement achieved by SVG outperforms other data augmentation algorithms. Further results from the ablation study verify the effectiveness of each component of SVG. Finally, the studies in the control group with varying numbers of samples show that the SVG algorithm consistently improves the AUC, with improvements ranging from approximately 0.02 to 0.15.

Convergence properties of data augmentation algorithms for high-dimensional robit regression

The logistic and probit link functions are the most common choices for regression models with a binary response. However, these choices are not robust to the presence of outliers/unexpected observations. The robit link function, which is equal to the inverse CDF of the Student’s t-distribution, provides a robust alternative to the probit and logistic link functions. A multivariate normal prior for the regression coefficients is the standard choice for Bayesian inference in robit regression models. The resulting posterior density is intractable and a Data Augmentation (DA) Markov chain is used to generate approximate samples from the desired posterior distribution. Establishing geometric ergodicity for this DA Markov chain is important as it provides theoretical guarantees for asymptotic validity of MCMC standard errors for desired posterior expectations/quantiles. Previous work [16] established geometric ergodicity of this robit DA Markov chain assuming (i) the sample size n dominates the number of predictors p, and (ii) an additional constraint which requires the sample size to be bounded above by a fixed constant which depends on the design matrix X. In particular, modern high-dimensional settings where n<p are not considered. In this work, we show that the robit DA Markov chain is trace-class (i.e., the eigenvalues of the corresponding Markov operator are summable) for arbitrary choices of the sample size n, the number of predictors p, the design matrix X, and the prior mean and variance parameters. The trace-class property implies geometric ergodicity. Moreover, this property allows us to conclude that the sandwich robit chain (obtained by inserting an inexpensive extra step in between the two steps of the DA chain) is strictly better than the robit DA chain in an appropriate sense, and enables the use of recent methods to estimate the spectral gap of trace class DA Markov chains.

Strong Generalized Speech Emotion Recognition Based on Effective Data Augmentation

The absence of labeled samples limits the development of speech emotion recognition (SER). Data augmentation is an effective way to address sample sparsity. However, there is a lack of research on data augmentation algorithms in the field of SER. In this paper, the effectiveness of classical acoustic data augmentation methods in SER is analyzed, based on which a strong generalized speech emotion recognition model based on effective data augmentation is proposed. The model uses a multi-channel feature extractor consisting of multiple sub-networks to extract emotional representations. Different kinds of augmented data that can effectively improve SER performance are fed into the sub-networks, and the emotional representations are obtained by the weighted fusion of the output feature maps of each sub-network. And in order to make the model robust to unseen speakers, we employ adversarial training to generalize emotion representations. A discriminator is used to estimate the Wasserstein distance between the feature distributions of different speakers and to force the feature extractor to learn the speaker-invariant emotional representations by adversarial training. The simulation experimental results on the IEMOCAP corpus show that the performance of the proposed method is 2-9% ahead of the related SER algorithm, which proves the effectiveness of the proposed method.

Asynchronous and Distributed Data Augmentation for Massive Data Settings

Data augmentation (DA) algorithms are slow in massive data settings due to multiple passes through the entire data. We address this problem by developing a DA extension that exploits asynchronous and distributed computing. The extended DA algorithm is called Asynchronous and Distributed (AD) DA with the original DA as its parent. Any ADDA is indexed by a parameter r ∈ ( 0 , 1 ) and starts by dividing the entire data into k disjoint subsets and storing them on k processes. Every iteration of ADDA augments only an r-fraction of the k data subsets with some positive probability and leaves the remaining ( 1 − r ) -fraction of the augmented data unchanged. The parameter draws are obtained using the r-fraction of new and ( 1 − r ) -fraction of old augmented data. We show that the ADDA Markov chain is Harris ergodic with the desired stationary distribution under mild conditions on the parent DA algorithm. We demonstrate that ADDA is significantly faster than its parent for many (k, r) choices in three representative models. We also establish the geometric ergodicity of the ADDA Markov chain for all the three models, which yields asymptotically valid standard errors for estimates of desired posterior quantities. Supplementary materials for this article are available online.

A Method for Analyzing the Performance Impact of Imbalanced Binary Data on Machine Learning Models

Machine learning models may not be able to effectively learn and predict from imbalanced data in the fields of machine learning and data mining. This study proposed a method for analyzing the performance impact of imbalanced binary data on machine learning models. It systematically analyzes 1. the relationship between varying performance in machine learning models and imbalance rate (IR); 2. the performance stability of machine learning models on imbalanced binary data. In the proposed method, the imbalanced data augmentation algorithms are first designed to obtain the imbalanced dataset with gradually varying IR. Then, in order to obtain more objective classification results, the evaluation metric AFG, arithmetic mean of area under the receiver operating characteristic curve (AUC), F-measure and G-mean are used to evaluate the classification performance of machine learning models. Finally, based on AFG and coefficient of variation (CV), the performance stability evaluation method of machine learning models is proposed. Experiments of eight widely used machine learning models on 48 different imbalanced datasets demonstrate that the classification performance of machine learning models decreases with the increase of IR on the same imbalanced data. Meanwhile, the classification performances of LR, DT and SVC are unstable, while GNB, BNB, KNN, RF and GBDT are relatively stable and not susceptible to imbalanced data. In particular, the BNB has the most stable classification performance. The Friedman and Nemenyi post hoc statistical tests also confirmed this result. The SMOTE method is used in oversampling-based imbalanced data augmentation, and determining whether other oversampling methods can obtain consistent results needs further research. In the future, an imbalanced data augmentation algorithm based on undersampling and hybrid sampling should be used to analyze the performance impact of imbalanced binary data on machine learning models.

Flexible and high quality plant growth prediction with limited data.

Predicting plant growth is a fundamental challenge that can be employed to analyze plants and further make decisions to have healthy plants with high yields. Deep learning has recently been showing its potential to address this challenge in recent years, however, there are still two issues. First, image-based plant growth prediction is currently taken either from time series or image generation viewpoints, resulting in a flexible learning framework and clear predictions, respectively. Second, deep learning-based algorithms are notorious to require a large-scale dataset to obtain a competing performance but collecting enough data is time-consuming and expensive. To address the issues, we consider the plant growth prediction from both viewpoints with two new time-series data augmentation algorithms. To be more specific, we raise a new framework with a length-changeable time-series processing unit to generate images flexibly. A generative adversarial loss is utilized to optimize our model to obtain high-quality images. Furthermore, we first recognize three key points to perform time-series data augmentation and then put forward T-Mixup and T-Copy-Paste. T-Mixup fuses images from a different time pixel-wise while T-Copy-Paste makes new time-series images with a different background by reusing individual leaves extracted from the existing dataset. We perform our method in a public dataset and achieve superior results, such as the generated RGB images and instance masks securing an average PSNR of 27.53 and 27.62, respectively, compared to the previously best 26.55 and 26.92.

Generative adversarial neural networks augment marker and pathway analysis of treatment resistant HPV+ head and neck squamous cell carcinoma.

e18059 Background: Biological behaviors of cancers are extremely complex and governed via nonlinear interplay of thousands of genes. This complexity has become a major hurdle to generating mechanistic conclusions in rare clinical scenarios, such as treatment resistant HPV+ head and neck squamous cell carcinoma, when observations are limited. Recently, data augmentation algorithms have served to increase the number and diversity of training observations while preserving the underlying structure and nonlinear relationships of data features. Deep learning techniques such as generative adversarial neural networks (GANs) have shown promise as robust tools for data augmentation of nonimage data. This study aims to evaluate the utility of employing separate GANs to highlight gene expression differences between treatment sensitive (TS) vs treatment resistant (TR) HPV+ head and neck squamous cell carcinoma. Methods: We performed RNAseq on 17 patients with HPV-positive oropharyngeal squamous cell carcinoma (OPSCC). We defined TR patients as those with locally advanced OPSCC at presentation, who rapidly recurred following receipt of standard of care therapy (n = 11). In most cases, patients failed to respond to salvage therapies and rapidly died. In contrast, we defined TS patients as those with locally advanced OPSCC at presentation, who received standard of care therapy, and were followed for 5 years without any evidence of recurrence (n = 6). These two cohorts were used to train two separate GANs, which synthesized 500 robust synthetic samples of each phenotype for downstream analysis. Results: Analysis of 17 samples yielded less than 100 differentially expressed (DE) genes passing FDR cutoffs, inhibiting pathway enrichment analysis (PAE). DE testing between synthetic samples yielded 1196 DE genes. Validating our deep learning approach, these genes belong to coherent pathways as determined by PAE. Namely, the VEGFA-VEGFR2 pathway was strongly enriched in the TR population (p = 1.5 x 10-4). Further, the NKT cell signature (p = 3.0x10-62), and the gene library of T-cell receptor induction of apoptosis (p = 5.4 x 10-9), was enriched in the TS population. Lastly, to validate the experimental method, DE testing on GAN augmented cohorts of renal papillary vs renal clear cell samples from The Cancer Genome Atlas, trained on 10 samples per condition, reliably predicted DE genes when comparing the full cohort of renal papillary n = 534 vs renal clear cell n = 295 samples, with an accuracy of 93% and F1 score of 88%. Conclusions: To our knowledge this is the first study to utilize separate GANs to augment structural gene expression changes occurring in rare clinical scenarios such as treatment resistant HPV+ OPSCC. Utilizing this method uncovers changes in angiogenesis and T-cell function between treatment sensitive and resistant HPV+ OPSCC.

Modified Pólya-Gamma data augmentation for Bayesian analysis of directional data

In this work, we develop new data augmentation algorithms for Bayesian analysis of directional data using the von Mises-Fisher distribution in arbitrary dimensions. The approach leads to a new class of distributions, called the Modified Pólya-Gamma distribution, which we construct in detail. The proposed data augmentation strategies circumvent the need for analytic approximations to integration, numerical integration, or Metropolis-Hastings for the corresponding posterior inference. Simulations and real data examples are presented to demonstrate the applicability and to apprise the performance of the proposed procedures.

End-to-End Structural analysis in civil engineering based on deep learning

This paper presents DeepSNA (Deep Structural Nonlinear Analysis), the first general end-to-end computational framework in civil engineering that can predict the full range of mechanical responses of different structures based on deep learning. The proposed framework comprehensively considers intrinsic structural information and external excitations from both the data and the model. First, a data interface schema was carefully designed to eliminate manual interventions. Based on the distinctive characteristics of structural analysis, we proposed deep learning models for static and dynamic features that could identify underlying mechanical coupling relationships and historical dependencies. Moreover, we designed data augmentation algorithms to address the lack of data in real-world applications. Finally, we developed the DeepSNA framework and validated it with steel plate shear wall structures. The results indicated that the new framework was more accurate and had at least 1000 times greater computational efficiency than conventional numerical methods.

Survey on Videos Data Augmentation for Deep Learning Models

In most Computer Vision applications, Deep Learning models achieve state-of-the-art performances. One drawback of Deep Learning is the large amount of data needed to train the models. Unfortunately, in many applications, data are difficult or expensive to collect. Data augmentation can alleviate the problem, generating new data from a smaller initial dataset. Geometric and color space image augmentation methods can increase accuracy of Deep Learning models but are often not enough. More advanced solutions are Domain Randomization methods or the use of simulation to artificially generate the missing data. Data augmentation algorithms are usually specifically designed for single images. Most recently, Deep Learning models have been applied to the analysis of video sequences. The aim of this paper is to perform an exhaustive study of the novel techniques of video data augmentation for Deep Learning models and to point out the future directions of the research on this topic.

State of Health Estimation of Lithium-Ion Battery Using Time Convolution Memory Neural Network

The accurate state of health (SOH) estimation of lithium-ion batteries enables users to make wise replacement decision and reduce economic losses. SOH estimation accuracy is related to many factors, such as usage time, ambient temperature, charge and discharge rate, etc. Thus, proper extraction of features from the above factors becomes a great challenge. In order to extract battery’s features effectively and improve SOH estimation accuracy, this article proposes a time convolution memory neural network (TCMNN), combining convolutional neural networks (CNN) and long short-term memory (LSTM) by dropout regularization-based fully connected layer. In experiment, the terminal voltage and charging current of the battery during charging process are collected, and input and output data sets are sorted out from the experimental battery data. Due to the limited equipment in the laboratory, only one battery can be charged and discharged at a time; the amount of battery data collected is relatively small, which will affect the extraction of features during the training process. Data augmentation algorithms are applied to solve the problem. Furthermore, in order to improve the accuracy of estimation, exponential smoothing algorithm is used to optimize output data. The results show that the proposed method can well extract and learn the feature relationship of battery cycle charge and discharge process in a long time span. In addition, it has higher accuracy than that of CNN, LSTM, Backpropagation (BP) algorithm, and Grey model-based neural network. The maximum error is limited to 3.79%, and the average error is limited to 0.143%, while the input data dimension is 514.

The Learning Curve of Artificial Intelligence for Dental Implant Treatment Planning: A Descriptive Study

Introduction: Cone-beam computed tomography (CBCT) has been applied to implant dentistry. The increasing use of this technology produces a critical number of images that can be used for training artificial intelligence (AI). Objectives: To investigate the learning curve of the developed AI for dental implant planning in the posterior maxillary region. Methods: A total of 184 CBCT image sets of patients receiving posterior maxillary implants were processed with software (DentiPlan Pro version 3.7; NECTEC, NSTDA, Thailand) to acquire 316 implant position images. The planning software image interfaces were anonymously captured with full-screen resolution. Three hundred images were randomly sorted to create six data sets, including 1–50, 1–100, 1–150, 1–200, 1–250, and 1–300. The data sets were used to develop AI for dental implant planning through the IBM PowerAI Vision platform (IBM Thailand Co., Ltd., Bangkok, Thailand) by using a faster R-CNN algorithm. Four data augmentation algorithms, including blur, sharpen, color, and noise, were also integrated to observe the improvement of the model. After the testing process with 16 images that were not included in the training set, the recorded data were analyzed for detection and accuracy to generate the learning curve of the model. Results: The learning curve revealed some similar patterns. The curve trend of the original and blurred augmented models was in a similar pattern in the panoramic image. In the last training set, the blurred augmented model improved the detection by 12.50%, but showed less accuracy than the original model by 18.34%, whereas the other three augmented models had different patterns. They were continuously increasing in both detection and accuracy. However, their detection dropped in the last training set. The colored augmented model demonstrated the best improvement with 40% for the panoramic image and 18.59% for the cross-sectional image. Conclusions: Within the limitation of the study, it may be concluded that the number of images used in AI development is positively related to the AI interpretation. The data augmentation techniques to improve the ability of AI are still questionable.

Audiogmenter: a MATLAB toolbox for audio data augmentation

Purpose Create and share a MATLAB library that performs data augmentation algorithms for audio data. This study aims to help machine learning researchers to improve their models using the algorithms proposed by the authors. Design/methodology/approach The authors structured our library into methods to augment raw audio data and spectrograms. In the paper, the authors describe the structure of the library and give a brief explanation of how every function works. The authors then perform experiments to show that the library is effective. Findings The authors prove that the library is efficient using a competitive dataset. The authors try multiple data augmentation approaches proposed by them and show that they improve the performance. Originality/value A MATLAB library specifically designed for data augmentation was not available before. The authors are the first to provide an efficient and parallel implementation of a large number of algorithms.

Intelligent rotating machinery fault diagnosis based on super-resolution enhancement using data augmentation under large speed fluctuation

In the real production of industry, in order to solve the problem that it is usually difficult to obtain correctly labeled samples, data augmentation algorithms have received more and more attention. Many efficient deep learning models have been successfully applied to the intelligent fault diagnosis of rotating machinery. However, the premise of the above method is the working conditions of the machinery are constant. It is inevitable that the equipment runs under large speed fluctuation in real industries. To achieve data augmentation under the condition of variable speed, an efficient sub-pixel deep residual neural network (ESPDRN) is proposed. The ESPDRN framework is implemented as follows: utilize low resolution (LR) samples as input to the network, to effectively learn high-level and extract feature maps by constructing a deep residual network. Through sub-pixel convolution layers to arrange the LR features of multi-channels periodically and get a set of ultra high resolution features, and the data points are augmented 16 times compared to raw data. Statistical indicators and experimental results demonstrate that the proposed ESPDRN model can effectively generate data under the condition of variable speed.

Computer tomographic differential diagnosis of ameloblastoma and odontogenic keratocyst: classification using a convolutional neural network.

To analyse the automatic classification performance of a convolutional neural network (CNN), Google Inception v3, using tomographic images of odontogenic keratocysts (OKCs) and ameloblastomas (AMs). For construction of the database, we selected axial multidetector CT images from patients with confirmed AM (n = 22) and OKC (n = 18) based on a conclusive histopathological report. The images (n = 350) were segmented manually and data augmentation algorithms were applied, totalling 2500 images. The k-fold × five cross-validation method (k = 2) was used to estimate the accuracy of the CNN model. The accuracy and standard deviation (%) of cross-validation for the five iterations performed were 90.16 ± 0.95, 91.37 ± 0.57, 91.62 ± 0.19, 92.48 ± 0.16 and 91.21 ± 0.87, respectively. A higher error rate was observed for the classification of AM images. This study demonstrated a high classification accuracy of Google Inception v3 for tomographic images of OKCs and AMs. However, AMs images presented the higher error rate.

Novel Data Augmentation Employing Multivariate Gaussian Distribution for Neural Network-Based Blood Pressure Estimation

In this paper, we propose a novel data augmentation technique employing multivariate Gaussian distribution (DA-MGD) for neural network (NN)-based blood pressure (BP) estimation, which incorporates the relationship between the features in a multi-dimensional feature vector to describe the correlated real-valued random variables successfully. To verify the proposed algorithm against the conventional algorithm, we compare the results in terms of mean error (ME) with standard deviation and Pearson correlation using 110 subjects contributed to the database (DB) which includes the systolic BP (SBP), diastolic BP (DBP), photoplethysmography (PPG) signal, and electrocardiography (ECG) signal. For each subject, 3 times (or 6 times) measurements are accomplished in which the PPG and ECG signals are recorded for 20 s. And, to compare with the performance of the BP estimation (BPE) using the data augmentation algorithms, we train the BPE model using the two-stage system, called the stacked NN. Since the proposed algorithm can express properly the correlation between the features than the conventional algorithm, the errors turn out lower compared to the conventional algorithm, which shows the superiority of our approach.

Differential diagnosis of ameloblastoma and odontogenic keratocyst by machine learning of panoramic radiographs

PurposeThe differentiation of the ameloblastoma and odontogenic keratocyst directly affects the formulation of surgical plans, while the results of differential diagnosis by imaging alone are not satisfactory. This paper aimed to propose an algorithm based on convolutional neural networks (CNN) structure to significantly improve the classification accuracy of these two tumors.MethodsA total of 420 digital panoramic radiographs provided by 401 patients were acquired from the Shanghai Ninth People’s Hospital. Each of them was cropped to a patch as a region of interest by radiologists. Furthermore, inverse logarithm transformation and histogram equalization were employed to increase the contrast of the region of interest (ROI). To alleviate overfitting, random rotation and flip transform as data augmentation algorithms were adopted to the training dataset. We provided a CNN structure based on a transfer learning algorithm, which consists of two branches in parallel. The output of the network is a two-dimensional vector representing the predicted scores of ameloblastoma and odontogenic keratocyst, respectively.ResultsThe proposed network achieved an accuracy of 90.36% (AUC = 0.946), while sensitivity and specificity were 92.88% and 87.80%, respectively. Two other networks named VGG-19 and ResNet-50 and a network trained from scratch were also used in the experiment, which achieved accuracy of 80.72%, 78.31%, and 69.88%, respectively.ConclusionsWe proposed an algorithm that significantly improves the differential diagnosis accuracy of ameloblastoma and odontogenic keratocyst and has the utility to provide a reliable recommendation to the oral maxillofacial specialists before surgery.

MeshCut data augmentation for deep learning in computer vision.

To solve overfitting in machine learning, we propose a novel data augmentation method called MeshCut, which uses a mesh-like mask to segment the whole image to achieve more partial diversified information. In our experiments, this strategy outperformed the existing augmentation strategies and achieved state-of-the-art results in a variety of computer vision tasks. MeshCut is also an easy-to-implement strategy that can efficiently improve the performance of the existing convolutional neural network models by a good margin without careful hand-tuning. The performance of such a strategy can be further improved by incorporating it into other augmentation strategies, which can make MeshCut a promising baseline strategy for future data augmentation algorithms.

Parameter-expanded data augmentation for analyzing correlated binary data using multivariate probit models.

Data augmentation has been commonly utilized to analyze correlated binary data using multivariate probit models in Bayesian analysis. However, the identification issue in the multivariate probit models necessitates a rigorous Metropolis-Hastings algorithm for sampling a correlation matrix, which may cause slow convergence and inefficiency of Markov chains. It is well-known that the parameter-expanded data augmentation, by introducing a working/artificial parameter or parameter vector, makes an identifiable model be non-identifiable and improves the mixing and convergence of data augmentation components. Therefore, we motivate to develop efficient parameter-expanded data augmentations to analyze correlated binary data using multivariate probit models. We investigate both the identifiable and non-identifiable multivariate probit models and develop the corresponding parameter-expanded data augmentation algorithms. We point out that the approaches, based on one non-identifiable model, circumvent a Metropolis-Hastings algorithm for sampling a correlation matrix and improve the convergence and mixing of correlation parameters; the identifiable model may produce the estimated regression parameters with smaller standard errors than the non-identifiable model does. We illustrate our proposed approaches using simulation studies and through the application to a longitudinal dataset from the Six Cities study.

LDA-based data augmentation algorithm for acoustic scene classification

Deep neural network needs large amount of data for training, to obtain more data, many simple data augmentation algorithms have been proposed. In this paper, we propose a LDA-based data augmentation algorithm to extend the training set. The proposed LDA-based data augmentation algorithm uses the topic model LDA to detect the key audio words in the recordings, and further to detect the key audio events and non-key audio events for each recording; with the detected key-audio-event segments, for each acoustic scene class, the probability distribution of key-audio-event’s occurrence numbers, the probability distribution of key-audio-event’s locations under each occurrence number and the probability distribution of key-audio-event’s durations under each occurrence number is counted, and then the new recordings are generated according to these probability distributions. Experiments are done on the public TUT acoustic scenes 2016 dataset, and the experimental results show that compared with the other simple data augmentation algorithms, the proposed LDA-based data augmentation algorithm is more stable and effective, it can get better generalization ability for different kinds of neural network on different datasets.