Multi-Class Ground Truth Inference in Crowdsourcing with Clustering

Noise correction to improve data and model quality for crowdsourcing

In crowdsourcing scenarios, we can obtain each instance’s multiple noisy labels set from different crowd workers and then use a ground truth inference algorithm to infer its integrated label. Despite the effectiveness of ground truth inference algorithms, a certain level of noise still remains in the integrated labels. To reduce the impact of noise, many noise correction algorithms have been proposed in recent years. To the best of our knowledge, however, nearly all existing noise correction algorithms only exploit each instance’s own multiple noisy label sets but ignore the multiple noisy label sets of its neighbors. Here neighbors refer to the nearest instances found in the feature space based on the distance metric learning. In this article, we propose neighborhood weighted voting-based noise correction (NWVNC). In NWVNC, we at first take advantage of the multiple noisy label sets of each instance’s neighbors (including itself) to estimate the probability that it belongs to its integrated label. Then, we use the estimated probability to identify and filter noise instances and thus obtain a clean set and a noise set. Finally, we train three heterogeneous classifiers on the clean set and correct the noise instances by the consensus voting of three trained classifiers. The experimental results on 34 simulated and two real-world crowdsourced datasets show that NWVNC significantly outperforms all the other state-of-the-art noise correction algorithms used for comparison.

Three-way decision-based label integration for crowdsourcing

With the development of crowdsourcing, data acquisition for supervised learning from annotators all over the world becomes simple and economical. To improve accuracy, it is nature to obtain multiple noisy labels (i.e., a multiple label set) for each example from the crowd. Then, consensus algorithms can infer the estimated ground truth from the multiple label set for each example. The estimated ground truth is also called an integrated label, which could be a noise. That is, a dataset constructed via integrating the multiple noisy labels for each example in a crowdsourcing dataset (called an integrated dataset) still contains noises. In order to further improve the data quality of an integrated dataset, so that to improve the performance of a model learned from the integrated dataset, this paper proposes a framework that integrates active learning with the self-healing of a model together. With active learning, a limited number of examples from the integrated dataset, which are most likely noises, are selected for the oracle to correct; with the self-healing of a model, the data quality of the integrated dataset can be also improved automatically. From our experimental results on eight simulated crowdsourcing datasets with three popular consensus algorithms, we draw some conclusions as follows. (1) Our proposed framework does improve the performance of a model learned from the integrated dataset. (2) The simple active learning selection strategy based on uncertainty estimation can identify noises in the integrated dataset. (3) Self-healing is efficient and effective to improve the data quality of the integrated dataset, so that it improves the accuracy of a model learned from the integrated dataset. We further investigate our proposed framework on a real-world crowdsourcing dataset collected from Amazon Mechanical Turk, and the above conclusions are sustained.

Worker similarity-based noise correction for crowdsourcing

Improving Bayesian credibility intervals for classifier error rates using maximum entropy empirical priors

Instance difficulty-based noise correction for crowdsourcing

Point cloud segmentation is a fundamental task in 3D. Despite recent progress on point cloud segmentation with the power of deep networks, current learning methods based on the clean label assumptions may fail with noisy labels. Yet, class labels are often mislabeled at both instance-level and boundary-level in real-world datasets. In this work, we take the lead in solving the instance-level label noise by proposing a Point Noise-Adaptive Learning (PNAL) framework. Compared to noise-robust methods on image tasks, our framework is noise-rate blind, to cope with the spatially variant noise rate specific to point clouds. Specifically, we propose a point-wise confidence selection to obtain reliable labels from the historical predictions of each point. A cluster-wise label correction is proposed with a voting strategy to generate the best possible label by considering the neighbor correlations. To handle boundary-level label noise, we also propose a variant "PNAL-boundary" with a progressive boundary label cleaning strategy. Extensive experiments demonstrate its effectiveness on both synthetic and real-world noisy datasets. Even with 60% symmetric noise and high-level boundary noise, our framework significantly outperforms its baselines, and is comparable to the upper bound trained on completely clean data. Moreover, we cleaned the popular real-world dataset ScanNetV2 for rigorous experiment. Our code and data is available at https://github.com/pleaseconnectwifi/PNAL.

Machine learning for Big Data analytics in plants.

Traditional truth inference algorithms take multiple source labels as input and infer true labels for objects. Besides source labels, object features have been introduced in inference algorithms to achieve superior performance. A typical algorithm such as learning from crowds learns a classification model with the guide of inferred true labels where true labels are inferred from source labels. However, the main shortcoming exists in current algorithms and limits their inference performance: label noise. Since source labels from real-world data are noisy, a classifier is likely to be misguided to learn an imprecise decision boundary. In this paper, we propose a deep clustering-based aggregation model (DCAM) to overcome the shortcoming. DCAM introduces clustering for object features to form fine-grained clusters, where objects in the same cluster are supposed to have similar labels. DCAM exploits a cluster label distribution to represent the labeling information of all objects in the corresponding cluster to overcome the problem of label noise. To implement the idea of clustering-based truth inference, DCAM integrates source label generation and deep clustering in a unified framework by utilizing maximum a posteriori (MAP) estimation. Therefore, the proposed model is a novel approach for truth inference with object features. Experimental results on eight real-world inference tasks show that DCAM has a significant improvement of inference accuracy over the state-of-the-art truth inference algorithms. We further discuss the effect of cluster numbers, the quality of clustering, and illustrate the learned embeddings to support the effectiveness of DCAM.

Multi label classification is concerned with learning from a set of instances that are associated with a set of labels, that is, an instance could be associated with multiple labels at the same time. This task occurs frequently in application areas like text categorization, multimedia classification, bioinformatics, protein function classification and semantic scene classification. Current multi-label classification methods could be divided into two categories. The first is called problem transformation methods, which transform multi-label classification problem into single label classification problem, and then apply any single label classifier to solve the problem. The second category is called algorithm adaptation methods, which adapt an existing single label classification algorithm to handle multi-label data. In this paper, we propose a multi-label classification approach based on correlations among labels that use both problem transformation methods and algorithm adaptation methods. The approach begins with transforming multi-label dataset into a single label dataset using least frequent label criteria, and then applies the PART algorithm on the transformed dataset. The output of the approach is multi-labels rules. The approach also tries to get benefit from positive correlations among labels using predictive Apriori algorithm. The proposed approach has been evaluated using two multi-label datasets named (Emotions and Yeast) and three evaluation measures (Accuracy, Hamming Loss, and Harmonic Mean). The experiments showed that the proposed approach has a fair accuracy in comparison to other related methods.

Training with noisy class labels impairs neural networks’ generalization performance. In this context, mixup is a popular regularization technique to improve training robustness by making memorizing false class labels more difficult. However, mixup neglects that multiple annotators, e.g., crowdworkers, typically provide class labels. Therefore, we propose an extension of mixup, which handles multiple class labels per instance while considering which class label originates from which annotator. Integrated into our multi-annotator classification framework annot-mix, it performs superiorly to eleven (mostly state-of-the-art) approaches in an evaluation study with eleven datasets comprising noisy class labels from either human or simulated annotators. Our code is publicly available through our GitHub repository at https://github.com/ies-research/multi-annotator-machine-learning/tree/annot-mix.

In a conventional supervised learning paradigm, each data instance is associated with one single class label. Multilabel learning differs in the way that data instances may belong to multiple concepts simultaneously, which naturally appear in a variety of high impact domains, ranging from bioinformatics and information retrieval to multimedia analysis. It targets leveraging the multiple label information of data instances to build a predictive learning model that can classify unlabeled instances into one or multiple predefined target classes. In multilabel learning, even though each instance is associated with a rich set of class labels, the label information could be noisy and incomplete as the labeling process is both time consuming and labor expensive, leading to potential missing annotations or even erroneous annotations. The existence of noisy and missing labels could negatively affect the performance of underlying learning algorithms. More often than not, multilabeled data often has noisy, irrelevant, and redundant features of high dimensionality. The existence of these uninformative features may also deteriorate the predictive power of the learning model due to the curse of dimensionality. Feature selection, as an effective dimensionality reduction technique, has shown to be powerful in preparing high-dimensional data for numerous data mining and machine-learning tasks. However, a vast majority of existing multilabel feature selection algorithms either boil down to solving multiple single-labeled feature selection problems or directly make use of the imperfect labels to guide the selection of representative features. As a result, they may not be able to obtain discriminative features shared across multiple labels. In this article, to bridge the gap between a rich source of multilabel information and its blemish in practical usage, we propose a novel noise-resilient multilabel informed feature selection framework (MIFS) by exploiting the correlations among different labels. In particular, to reduce the negative effects of imperfect label information in obtaining label correlations, we decompose the multilabel information of data instances into a low-dimensional space and then employ the reduced label representation to guide the feature selection phase via a joint sparse regression framework. Empirical studies on both synthetic and real-world datasets demonstrate the effectiveness and efficiency of the proposed MIFS framework.

In pattern recognition supervised learning algorithms has been used to model the relationships among the characteristics of a group of objects (or patterns) and their class labels. This model is build based on a set of training images that have been labelled by an expert in a specific problem domain. In general, the labelling process consists in defining the object's class label as well as demarcating the precise part of the image in which it is located the object of interest. In many computer vision applications, for example in automatic visual inspection, the labelling process may be a laborious task which requires an extensive work, it could even become impractical when a large number of training images are needed to train the pattern recognition system. This limitation has led to the development of new learning algorithms that allow some ambiguity or "weakness" in the way class labels are assigned. These algorithms are known as weakly supervised learning algorithms. In this paper we present an evaluation of different weakly supervised learning paradigms grouped according to the representation of the object of interest (in one or several characteristic vectors) associated to its class label (single or multiple label). We perform different experiments in the automatic visual inspection context, using a benchmark dataset of texture images with artificial defects. We test different weakly supervised learning algorithms. Results show that representing an object using multiple instances, presents better results in the identification of defective texture image.

Multi-label classification has many applications in the text categorization, biology and medical diagnosis, in which multiple class labels can be assigned to each training instance simultaneously. As it is often the case that there are relationships between the labels, extracting the existing relationships between the labels and taking advantage of them during the training or prediction phases can bring about significant improvements. In this paper, we have introduced positive, negative and hybrid relationships between the class labels for the first time and we have proposed a method to extract these relations for a multi-label classification task and consequently, to use them in order to improve the predictions made by a multi-label classifier. We have conducted extensive experiments to assess the effectiveness of the proposed method. The obtained results advocate the merits of the proposed method in improving the multi-label classification results.