Scarcity Of Training Data Research Articles

Conditional diffusion model-based generation of speckle patterns for digital image correlation

Generating speckle images is crucial for calibrating errors in digital image correlation (DIC) algorithms, alleviating the scarcity of training data for deep learning-based DIC algorithms, and even for the broader application of digital speckle methods. Traditional methods for speckle image synthesis struggle to accurately replicate the brightness variations in the foreground and background of realistic speckle images, and these methods come with high computational costs. Additionally, there needs to be more scientific and systematic evaluation methods for the quality of generated DIC speckle images. Our study presents a high-fidelity method for generating DIC speckle images with cracks based on a conditional diffusion model. We established evaluation metrics for generated DIC speckle images based on multiscale spatial and spectrum domains. We verified them on an open-source dataset of DIC speckle images depicting cracks in stone masonry walls. The conditional diffusion model excels at forward tasks (DIC crack image segmentation) and inverse tasks (DIC speckle image generation), with particular strength in the latter. The proposed spatial and multiscale spectrum domain metrics allow for the comprehensive and accurate evaluation of the alignment between the synthetic and natural speckle images. The conditional diffusion model for DIC speckle image generation and evaluation proposed in this paper has the potential to enhance further the accuracy and robustness of DIC technology in experimental mechanics and engineering applications.

Dynamic decoding and dual synthetic data for automatic correction of grammar in low-resource scenario.

Grammar error correction systems are pivotal in the field of natural language processing (NLP), with a primary focus on identifying and correcting the grammatical integrity of written text. This is crucial for both language learning and formal communication. Recently, neural machine translation (NMT) has emerged as a promising approach in high demand. However, this approach faces significant challenges, particularly the scarcity of training data and the complexity of grammar error correction (GEC), especially for low-resource languages such as Indonesian. To address these challenges, we propose InSpelPoS, a confusion method that combines two synthetic data generation methods: the Inverted Spellchecker and Patterns+POS. Furthermore, we introduce an adapted seq2seq framework equipped with a dynamic decoding method and state-of-the-art Transformer-based neural language models to enhance the accuracy and efficiency of GEC. The dynamic decoding method is capable of navigating the complexities of GEC and correcting a wide range of errors, including contextual and grammatical errors. The proposed model leverages the contextual information of words and sentences to generate a corrected output. To assess the effectiveness of our proposed framework, we conducted experiments using synthetic data and compared its performance with existing GEC systems. The results demonstrate a significant improvement in the accuracy of Indonesian GEC compared to existing methods.

ECEA: Extensible Co-Existing Attention for Few-Shot Object Detection.

Few-shot object detection (FSOD) identifies objects from extremely few annotated samples. Most existing FSOD methods, recently, apply the two-stage learning paradigm, which transfers the knowledge learned from abundant base classes to assist the few-shot detectors by learning the global features. However, such existing FSOD approaches seldom consider the localization of objects from local to global. Limited by the scarce training data in FSOD, the training samples of novel classes typically capture part of objects, resulting in such FSOD methods being unable to detect the completely unseen object during testing. To tackle this problem, we propose an Extensible Co-Existing Attention (ECEA) module to enable the model to infer the global object according to the local parts. Specifically, we first devise an extensible attention mechanism that starts with a local region and extends attention to co-existing regions that are similar and adjacent to the given local region. We then implement the extensible attention mechanism in different feature scales to progressively discover the full object in various receptive fields. In the training process, the model learns the extensible ability on the base stage with abundant samples and transfers it to the novel stage of continuous extensible learning, which can assist the few-shot model to quickly adapt in extending local regions to co-existing regions. Extensive experiments on the PASCAL VOC and COCO datasets show that our ECEA module can assist the few-shot detector to completely predict the object despite some regions failing to appear in the training samples and achieve the new state-of-the-art compared with existing FSOD methods. Code is released at https://github.com/zhimengXin/ECEA.

Multi-supervision transformer combining bounding box and mask for data-limited pose estimation

In scenarios where acquiring accurate annotated pose data is difficult, obtaining sufficient training samples becomes a challenge. Given these data constraints, it is challenging to provide optimal end-to-end supervision with limited pose information. To address this limitation, we propose a novel approach, named MSPose, which combines diverse data sources from human instances to provide multi-level supervision in human pose estimation. This approach mitigates the impact of scarce training data and enhances pose estimation precision by synergistically incorporating human bounding box and mask information. Our proposed method achieves multi-level model activation for the human body region, structure, and joints. This hierarchical pixel activation is facilitated by integrating human bounding box and mask signals as supplementary supervision during training. Importantly, we exclude auxiliary branches to ensure complete freedom during inference. Our experimental validation on a quarter of the COCO and MPII datasets demonstrates the competitiveness of our method compared to state-of-the-art methods. Compared to TokenPose-T, our MSPose-T achieves a 6.1 points improvement in average precision and a corresponding 3.1 points increase on the full COCO validation set. On the MPII dataset, MSPose achieves the highest Mean score of 1.9 points. Our approach achieves these improvements while maintaining consistent parameters and GFLOPs, highlighting the advantage of our approach.

PACT - Prediction of amyloid cross-interaction by threading

Amyloid proteins are often associated with the onset of diseases, including Alzheimer’s, Parkinson’s and many others. However, there is a wide class of functional amyloids that are involved in physiological functions, e.g., formation of microbial biofilms or storage of hormones. Recent studies showed that an amyloid fibril could affect the aggregation of another protein, even from a different species. This may result in amplification or attenuation of the aggregation process. Insight into amyloid cross-interactions may be crucial for better understanding of amyloid diseases and the potential influence of microbial amyloids on human proteins. However, due to the demanding nature of the needed experiments, knowledge of such interactions is still limited. Here, we present PACT (Prediction of Amyloid Cross-interaction by Threading) - the computational method for the prediction of amyloid cross-interactions. The method is based on modeling of a heterogeneous fibril formed by two amyloidogenic peptides. The resulting structure is assessed by the structural statistical potential that approximates its plausibility and energetic stability. PACT was developed and first evaluated mostly on data collected in the AmyloGraph database of interacting amyloids and achieved high values of Area Under ROC (AUC=0.88) and F1 (0.82). Then, we applied our method to study the interactions of CsgA - a bacterial biofilm protein that was not used in our in-reference datasets, which is expressed in several bacterial species that inhabit the human intestines - with two human proteins. The study included alpha-synuclein, a human protein that is involved in Parkinson’s disease, and human islet amyloid polypeptide (hIAPP), which is involved in type 2 diabetes. In both cases, PACT predicted the appearance of cross-interactions. Importantly, the method indicated specific regions of the proteins, which were shown to play a central role in both interactions. We experimentally confirmed the novel results of the indicated CsgA fragments interacting with hIAPP based on the kinetic characteristics obtained with the ThT assay. PACT opens the possibility of high-throughput studies of amyloid interactions. Importantly, it can work with fairly long protein fragments, and as a purely physicochemical approach, it relies very little on scarce training data. The tool is available as a web server at https://pact.e-science.pl/pact/. The local version can be downloaded from https://github.com/KubaWojciechowski/PACT.

Reconstructing Close Human Interactions from Multiple Views

This paper addresses the challenging task of reconstructing the poses of multiple individuals engaged in close interactions, captured by multiple calibrated cameras. The difficulty arises from the noisy or false 2D keypoint detections due to inter-person occlusion, the heavy ambiguity in associating keypoints to individuals due to the close interactions, and the scarcity of training data as collecting and annotating motion data in crowded scenes is resource-intensive. We introduce a novel system to address these challenges. Our system integrates a learning-based pose estimation component and its corresponding training and inference strategies. The pose estimation component takes multi-view 2D keypoint heatmaps as input and reconstructs the pose of each individual using a 3D conditional volumetric network. As the network doesn't need images as input, we can leverage known camera parameters from test scenes and a large quantity of existing motion capture data to synthesize massive training data that mimics the real data distribution in test scenes. Extensive experiments demonstrate that our approach significantly surpasses previous approaches in terms of pose accuracy and is generalizable across various camera setups and population sizes. The code is available on our project page: https://github.com/zju3dv/CloseMoCap.

Multi-task bioassay pre-training for protein-ligand binding affinity prediction.

Protein-ligand binding affinity (PLBA) prediction is the fundamental task in drug discovery. Recently, various deep learning-based models predict binding affinity by incorporating the three-dimensional (3D) structure of protein-ligand complexes as input and achieving astounding progress. However, due to the scarcity of high-quality training data, the generalization ability of current models is still limited. Although there is a vast amount of affinity data available in large-scale databases such as ChEMBL, issues such as inconsistent affinity measurement labels (i.e. IC50, Ki, Kd), different experimental conditions, and the lack of available 3D binding structures complicate the development of high-precision affinity prediction models using these data. To address these issues, we (i) propose Multi-task Bioassay Pre-training (MBP), a pre-training framework for structure-based PLBA prediction; (ii) construct a pre-training dataset called ChEMBL-Dock with more than 300k experimentally measured affinity labels and about 2.8M docked 3D structures. By introducing multi-task pre-training to treat the prediction of different affinity labels as different tasks and classifying relative rankings between samples from the same bioassay, MBP learns robust and transferrable structural knowledge from our new ChEMBL-Dock dataset with varied and noisy labels. Experiments substantiate the capability of MBP on the structure-based PLBA prediction task. To the best of our knowledge, MBP is the first affinity pre-training model and shows great potential for future development. MBP web-server is now available for free at: https://huggingface.co/spaces/jiaxianustc/mbp.

Algorithm selection using edge ML and case-based reasoning

In practical data mining, a wide range of classification algorithms is employed for prediction tasks. However, selecting the best algorithm poses a challenging task for machine learning practitioners and experts, primarily due to the inherent variability in the characteristics of classification problems, referred to as datasets, and the unpredictable performance of these algorithms. Dataset characteristics are quantified in terms of meta-features, while classifier performance is evaluated using various performance metrics. The assessment of classifiers through empirical methods across multiple classification datasets, while considering multiple performance metrics, presents a computationally expensive and time-consuming obstacle in the pursuit of selecting the optimal algorithm. Furthermore, the scarcity of sufficient training data, denoted by dimensions representing the number of datasets and the feature space described by meta-feature perspectives, adds further complexity to the process of algorithm selection using classical machine learning methods. This research paper presents an integrated framework called eML-CBR that combines edge edge-ML and case-based reasoning methodologies to accurately address the algorithm selection problem. It adapts a multi-level, multi-view case-based reasoning methodology, considering data from diverse feature dimensions and the algorithms from multiple performance aspects, that distributes computations to both cloud edges and centralized nodes. On the edge, the first-level reasoning employs machine learning methods to recommend a family of classification algorithms, while at the second level, it recommends a list of the top-k algorithms within that family. This list is further refined by an algorithm conflict resolver module. The eML-CBR framework offers a suite of contributions, including integrated algorithm selection, multi-view meta-feature extraction, innovative performance criteria, improved algorithm recommendation, data scarcity mitigation through incremental learning, and an open-source CBR module, reshaping research paradigms. The CBR module, trained on 100 datasets and tested with 52 datasets using 9 decision tree algorithms, achieved an accuracy of 94% for correct classifier recommendations within the top k=3 algorithms, making it highly suitable for practical classification applications.

Deep neural networks with adaptive solution space for inverse design of multilayer deep-etched grating

This article presents an inverse design technique of multilayer deep-etched grating (MDEG) using a deep neural network with adaptive solution space. MDEG is a key component in multispectral imaging, significantly affecting the quality of acquired multispectral images. Inverse design problems with one-to-many mappings from an optical response to multiple structures of MDEG have been addressed using machine learning techniques. Prior efforts, such as the tandem network and convolutional neural network (CNN), approached the problem from a modeling perspective. In this proposed method, a deep neural network is trained using cross-entropy loss for inverse design, predicting the probability distribution across the discretized space. The solution space is then formed by selecting the top K values with the highest probability, enabling evaluation of an optimal solution using the forward modeling network and avoiding the challenges of one-to-many mapping problems. Given the scarcity of optical training data, we explore training the network with reduced datasets of 30,000, 50,000 and 100,000 samples, demonstrating that the network performs well with a small amount of training data. Computational experiments conducted on the generated test dataset demonstrate that the proposed method improves in-band and out-of-band efficiencies by 26% and 12%, respectively, using only one-third of the required data.

Optimal Transport Posterior Alignment for Cross-lingual Semantic Parsing

Abstract Cross-lingual semantic parsing transfers parsing capability from a high-resource language (e.g., English) to low-resource languages with scarce training data. Previous work has primarily considered silver-standard data augmentation or zero-shot methods; exploiting few-shot gold data is comparatively unexplored. We propose a new approach to cross-lingual semantic parsing by explicitly minimizing cross-lingual divergence between probabilistic latent variables using Optimal Transport. We demonstrate how this direct guidance improves parsing from natural languages using fewer examples and less training. We evaluate our method on two datasets, MTOP and MultiATIS++SQL, establishing state-of-the-art results under a few-shot cross-lingual regime. Ablation studies further reveal that our method improves performance even without parallel input translations. In addition, we show that our model better captures cross-lingual structure in the latent space to improve semantic representation similarity.1

Enhancing Crop Classification Accuracy through Synthetic SAR-Optical Data Generation Using Deep Learning

Crop classification using remote sensing data has emerged as a prominent research area in recent decades. Studies have demonstrated that fusing synthetic aperture radar (SAR) and optical images can significantly enhance the accuracy of classification. However, a major challenge in this field is the limited availability of training data, which adversely affects the performance of classifiers. In agricultural regions, the dominant crops typically consist of one or two specific types, while other crops are scarce. Consequently, when collecting training samples to create a map of agricultural products, there is an abundance of samples from the dominant crops, forming the majority classes. Conversely, samples from other crops are scarce, representing the minority classes. Addressing this issue requires overcoming several challenges and weaknesses associated with the traditional data generation methods. These methods have been employed to tackle the imbalanced nature of training data. Nevertheless, they still face limitations in effectively handling minority classes. Overall, the issue of inadequate training data, particularly for minority classes, remains a hurdle that the traditional methods struggle to overcome. In this research, we explore the effectiveness of a conditional tabular generative adversarial network (CTGAN) as a synthetic data generation method based on a deep learning network, for addressing the challenge of limited training data for minority classes in crop classification using the fusion of SAR-optical data. Our findings demonstrate that the proposed method generates synthetic data with a higher quality, which can significantly increase the number of samples for minority classes, leading to a better performance of crop classifiers. For instance, according to the G-mean metric, we observed notable improvements in the performance of the XGBoost classifier of up to 5% for minority classes. Furthermore, the statistical characteristics of the synthetic data were similar to real data, demonstrating the fidelity of the generated samples. Thus, CTGAN can be employed as a solution for addressing the scarcity of training data for minority classes in crop classification using SAR–optical data.

Motor Imagery EEG Classification Based on a Weighted Multi-branch Structure Suitable for Multisubject Data.

Electroencephalogram (EEG) signal recognition based on deep learning technology requires the support of sufficient data. However, training data scarcity usually occurs in subject-specific motor imagery tasks unless multisubject data can be used to enlarge training data. Unfortunately, because of the large discrepancies between data distributions from different subjects, model performance could only be improved marginally or even worsened by simply training on multisubject data. This paper proposes a novel weighted multi-branch (WMB) structure for handling multisubject data to solve the problem, in which each branch is responsible for fitting a pair of source-target subject data and adaptive weights are used to integrate all branches or select branches with the largest weights to make the final decision. The proposed WMB structure was applied to six well-known deep learning models (EEGNet, Shallow ConvNet, Deep ConvNet, ResNet, MSFBCNN, and EEG_TCNet) and comprehensive experiments were conducted on EEG datasets BCICIV-2a, BCICIV-2b, high gamma dataset (HGD) and two supplementary datasets. Superior results against the state-of-the-art models have demonstrated the efficacy of the proposed method in subject-specific motor imagery EEG classification. For example, the proposed WMB_EEGNet achieved classification accuracies of 84.14%, 90.23%, and 97.81% on BCICIV-2a, BCICIV-2b and HGD, respectively. It is clear that the proposed WMB structure is capable to make good use of multisubject data with large distribution discrepancies for subject-specific EEG classification.

Exploring challenges and approaches for ChatGPT in multilingual and multisectoral contexts

ChatGPT, a transformer-based chatbot model, has gained significant attention for its ability to generate natural and coherent responses. However, the model still faces several challenges that limit its performance and applicability. This essay explores the current challenges of ChatGPT and proposes solutions to overcome them. The challenges identified include the scarcity of diverse and high-quality training data, coherence and topic transition issues in long conversations, and the risk of overfitting during training. To address these challenges, the essay proposes several solutions. Transfer learning techniques are suggested to improve model generalization by pre-training on a large corpus and fine-tuning on specific chatbot tasks. Regularization methods such as dropout and weight decay are recommended to prevent overfitting and improve generalization. The design of more effective evaluation metrics, including F1 score and human evaluation, is proposed to accurately assess the model's performance. Additionally, incorporating contextual information from previous conversation turns is explored to enhance coherence.The proposed solutions are evaluated through experiments and benchmarking. The results demonstrate promising improvements in the performance of ChatGPT, including enhanced coherence, better topic transition, reduced overfitting, and higher-quality generated responses. This research contributes to the advancement of chatbot models by addressing the challenges faced by ChatGPT. The proposed solutions offer practical strategies to improve the model's performance and applicability in real-world scenarios. The findings have implications for various industries that rely on chatbot technology, enabling them to provide more natural and coherent interactions with users.

A Waste Classification model in Low-illumination scenes based on ConvNeXt

Waste classification is an essential part of environmental pollution management in modern society. Object detection is an accurate and efficient way to classify waste, which is conducive to recycling resources. However, due to low object discriminability, existing waste classification models cannot classify waste in low-illumination scenes. A waste classification model, Dark-Waste, is designed to classify wastes in a low-illumination scenario. Firstly, to solve the scarcity of training data, an efficient and low-cost Illumination Conversion method is proposed to generate the low-light image. Secondly, the improved ConvNeXt network is combined with YOLOv5 to accurately and efficiently classify waste. Finally, we validated the model on a self-built dataset in real scenarios. The experimental results show that Dark-Waste achieves the best detection performance in low-illumination scenes. The Dark-Waste provides a new approach to waste management in complex environments and effectively contributes to the sustainable development of the urban ecological environment.

Automatic Speaking Assessment of Spontaneous L2 Finnish and Swedish

ABSTRACT The development of automated systems for evaluating spontaneous speech is desirable for L2 learning, as it can be used as a facilitating tool for self-regulated learning, language proficiency assessment, and teacher training programs. However, languages with fewer learners face challenges due to the scarcity of training data. Recent advancements in machine learning have made it possible to develop systems with a limited amount of target domain data. To this end, we propose automatic speaking assessment systems for spontaneous L2 speech in Finnish and Finland Swedish, comprising six machine learning models each, and report their performance in terms of statistical evaluation criteria.

Adaptation for Automated Drift Detection in Electromechanical Machine Monitoring.

Practical machine learning applications for streaming data can involve concept drift (the change in statistical properties of data over time), one-shot or few-shot learning (starting with only one or a few examples for each class), a scarcity of representative training data, and extreme verification latency (only the initial dataset has ground-truth labels). This work presents a framework for organizing signal processing and machine learning techniques to provide adaptive classification and drift detection. Nonintrusive load monitoring (NILM) serves as an ideal case study, as modern sensing solutions provide a wellspring of electromechanical data sources. There is a lack of training datasets that generalize across different load and fault scenarios. Accordingly, training must be accomplished with a limited set of data when deploying a NILM to a new power system. Also, loads can exhibit concept drift over time either due to faults or normal variation. NILM field data is used as an illustrative case study to demonstrate the proposed framework for adaptation and drift tracking.

Leveraging global binary masks for structure segmentation in medical images

Deep learning (DL) models for medical image segmentation are highly influenced by intensity variations of input images and lack generalization due to primarily utilizing pixels’ intensity information for inference. Acquiring sufficient training data is another challenge limiting models’ applications. Here, we proposed to leverage the consistency of organs’ anatomical position and shape information in medical images. We introduced a framework leveraging recurring anatomical patterns through global binary masks for organ segmentation. Two scenarios were studied: (1) global binary masks were the only input for the U-Net based model, forcing exclusively encoding organs’ position and shape information for rough segmentation or localization. (2) Global binary masks were incorporated as an additional channel providing position/shape clues to mitigate training data scarcity. Two datasets of the brain and heart computed tomography (CT) images with their ground-truth were split into (26:10:10) and (12:3:5) for training, validation, and test respectively. The two scenarios were evaluated using full training split as well as reduced subsets of training data. In scenario (1), training exclusively on global binary masks led to Dice scores of 0.77 ± 0.06 and 0.85 ± 0.04 for the brain and heart structures respectively. Average Euclidian distance of 3.12 ± 1.43 mm and 2.5 ± 0.93 mm were obtained relative to the center of mass of the ground truth for the brain and heart structures respectively. The outcomes indicated encoding a surprising degree of position and shape information through global binary masks. In scenario (2), incorporating global binary masks led to significantly higher accuracy relative to the model trained on only CT images in small subsets of training data; the performance improved by 4.3%–125.3% and 1.3%–48.1% for 1–8 training cases of the brain and heart datasets respectively. The findings imply the advantages of utilizing global binary masks for building models that are robust to image intensity variations as well as an effective approach to boost performance when access to labeled training data is highly limited.

Physics-regulated dynamic mode decomposition for natural gas pipeline flow

This study introduces a novel integration of dynamic mode decomposition (DMD) with physical regulations for natural gas pipeline flow. It aims to address the limitations of purely data-driven models and the importance of incorporating the physics of complex dynamic systems. By considering the mass conservation law, the proposed model ensures that the predictions generated via DMD with control adhere to the physical laws, resulting in a multi-objective optimization problem. To verify its performance, the proposed model was evaluated using real-world data of natural gas pipelines. The results demonstrate its superior accuracy and ability to avoid physically implausible predictions, particularly under data-limited conditions. Despite an increase in the overall computational cost by approximately 15%, the model achieved up to 50% error reduction with scarce training data, highlighting its robustness and effectiveness. This study represents a significant advancement in data-driven modeling techniques by fulfilling the critical need for accurate and reliable predictions that respect physical constraints, thus enhancing the interpretability and validity of the results.

Effective medical leaf identification using hybridization of GMM-CNN

Medical plants play a vital role in curing many diseases. These plants, along with their leaves, have medicinal values. If these leaves are identified appropriately, they can be chosen directly to have more significant relief from the disease. Therefore, the identification of these medical species is a challenging task. The ecologically motivated Convolutional Neural Networks (CNNs) have substantially contributed to computer visual research. This article introduces a unique approach to medical leaf identification based on the hybridization of the Gaussian mixture model and a Convolutional Neural Network (GMM-CNN). The experimentation is performed on the Flavia dataset and is carried out using benchmark evaluation metrics. The parameters like index volume, probability of random index, and global consistency error are evaluated. The Python simulation model is utilized for the evaluation of the proposed methodology. The hybrid technique combining GMM and CNN has considerable potential in medical leaf identification. The experimental findings indicate that the hybrid strategy exhibits superior performances. The methodology suggested in this study demonstrates exceptional levels of accuracy, precision, and recall when applied to a wide range of medical leaf categories. Moreover, integrating Gaussian Mixture Model (GMM) and Convolutional Neural Network (CNN) addresses concerns associated with a scarcity of training data by offering a more resilient structure for extracting features and performing classification. By combining the advantages of statistical modelling with deep learning, we develop a resilient and precise system that has the potential to enhance botanical research, medical diagnostics, and environmental monitoring applications.

D2LFS2Net: Multi‐class skin lesion diagnosis using deep learning and variance‐controlled Marine Predator optimisation: An application for precision medicine

AbstractIn computer vision applications like surveillance and remote sensing, to mention a few, deep learning has had considerable success. Medical imaging still faces a number of difficulties, including intra‐class similarity, a scarcity of training data, and poor contrast skin lesions, notably in the case of skin cancer. An optimisation‐aided deep learning‐based system is proposed for accurate multi‐class skin lesion identification. The sequential procedures of the proposed system start with preprocessing and end with categorisation. The preprocessing step is where a hybrid contrast enhancement technique is initially proposed for lesion identification with healthy regions. Instead of flipping and rotating data, the outputs from the middle phases of the hybrid enhanced technique are employed for data augmentation in the next step. Next, two pre‐trained deep learning models, MobileNetV2 and NasNet Mobile, are trained using deep transfer learning on the upgraded enriched dataset. Later, a dual‐threshold serial approach is employed to obtain and combine the features of both models. The next step was the variance‐controlled Marine Predator methodology, which the authors proposed as a superior optimisation method. The top features from the fused feature vector are classified using machine learning classifiers. The experimental strategy provided enhanced accuracy of 94.4% using the publicly available dataset HAM10000. Additionally, the proposed framework is evaluated compared to current approaches, with remarkable results.