Search

Principal component analysis network (PCANet), as an unsupervised shallow network, demonstrates noticeable effectiveness on datasets of various volumes. It carries a two-layer convolution with PCA as filter learning method, followed by a block-wise histogram post-processing stage. Following the structure of PCANet, extreme learning machine auto-encoder (ELM-AE) variants are employed to replace the PCA’s role, which come from extreme learning machine network (ELMNet) and hierarchical ELMNet. ELMNet emphasizes the importance of orthogonal projection while overlooking non-linearity. The latter introduces complex pre-processing to overcome drawback of non-linear ELM-AE. In this paper, we analyze intrinsic characteristics of ELM-AE variants and accordingly propose a regularized ELM-AE, which combines non-linearity learning capability and approximately orthogonal projection. Experiments on image classification show the effectiveness compared to supervised convolutional neural networks and related shallow networks on unsupervised feature learning.

Extreme learning machine (ELM) is a popular method in machine learning with extremely few parameters, fast learning speed and model efficiency. Unsupervised feature learning based ELM receives rising research focus. Recently the ELM auto-encoder (ELM-AE) was proposed for this task, which develops the ELM based compact feature learning without sacrificing elegant solution. Compared with ELM-AE and following $$\ell _1$$-regularized ELM-AE, we introduce a sparse Bayesian learning scheme into ELM-AE for better generalization capability. A parallel training strategy is also integrated to improve time-efficiency of multi-output sparse Bayesian learning. Furthermore, pruning hidden nodes for better performance and efficiency according to estimated variances of prior distribution of output weights is achieved. Experiments on several datasets verify the effectiveness and efficiency of our proposed ELM-AE for unsupervised feature learning, compared with PCA, NMF, ELM-AE and $$\ell _1$$-regularized ELM-AE.

With the development of deep neural networks, researchers have developed lots of algorithms related to face and achieved comparable results to human-level performance on several databases. However, few feature extraction models work well in the real world when the subject which is to be recognized has limited samples, for example, only one ID photo can be obtained before the face recognition task. To our best knowledge, there is no face database which contains ID photos and pictures from the real world for a subject simultaneously. To fill this gap, we collected 100 celebrities’ ID photos and their about 1000 stills or life pictures and formed a face database called FDID. Besides, we proposed a novel face recognition algorithm and evaluated it with this new database on the real-life videos.

Image Classification is one of the key computer vision tasks. Among numerous machine learning methods, we choose the Extreme Learning Machine (ELM) for our image classification applications. This thesis contributes to four aspects of ELM netwroks. From the view of efficient input data, we have designed handcrafted feature extraction method for smile images classification. From the perspective of the distribution of random weights between the input layer and hidden layer, we have proposed and proved the effectiveness of the sparse binary ELM. Inspired by the deep architecture of deep learning, we have extended the single layer to multiple layers of ELM to achieve better performance on large image classification datasets. Finally, from the point of target coding, we have introduced and evaluated different target coding methods for image classification. Doctor of Philosophy

Extreme Learning Machine (ELM) is a popular method in machine learning with extremely few parameters, fast learning speed and model efficiency. While a significant drawback is that ELM is restricted by its single-layer structure and prized analytic solution. If simply stacking more layers, analytic solution of ELM will be intractable. Then gradient-based optimization method is preferred and that results into normal neural networks. Recently a multi-layer ELM (ML-ELM) is proposed to learn compact feature with a series of ELM auto-encoders, which attempts to extend ELM to a deeper network without sacrificing elegant solution. Compared with ML-ELM and following hierarchical ELM, we introduce a sparse Bayesian learning method to imply a stronger sparse regularization and prune network structure. Experiments on classification verify the efficiency of our proposed new multi-layer ELM for unsupervised feature learning.

Visual aided guide system (VaGS) is essential and important for the future UAVs, also for the mixed ground command and guide of both UAVs and manned fighters. The key module of VaGS is to recognize the commander‘s gestures and translate them into instructions. This paper introduces a new high-performance gesture recognition architecture for VaGS. It includes two main components: (1) a multiscale structure is adopted to conduct feature learning, which combined global human gesture features with local hands gesture features rather than an extremely deep neural networks; (2) the fused features are input into an extreme learning machine (ELM) for a classification procedure and then output the gesture instructions. Experimental results show that the proposed method can achieve a rather better expression in some standard datasets, and obtain an accuracy of 99.6% in an aviation guide gesture dataset with 40 classes built by ourselves, still a less training and forward time consumption.

Target coding is an indispensable part of supervised learning. Currently, the assumption of the most popular target coding like one-hot (one-of-K) coding is that targets are independent. However, this assumption is limited due to the complex relationship between targets. In this paper, we will explore the effects of kinds of target coding methods on the performance of Extreme Learning Machine Classifiers (ELM-C). Linearly independent coding (e.g., one-of-k coding, Hadamard coding) and linearly dependent coding (e.g., ordinal coding, binary coding) are analyzed and compared. The experimental results on OCR letter dataset show that different target coding will indeed affect the performance of the same classifier.

With the development of deep neural networks, researchers have developed lots of algorithms related to face and achieved comparable results to human-level performance on several databases. However, few feature extraction models work well in the real world when the subject which is to be recognized has limited samples, for example, only one ID photo can be obtained before the face recognition task. To our best knowledge, there is no face database which contains ID photos and pictures from the real world for a subject simultaneously. To fill this gap, we collected 100 celebrities’ ID photos and their about 1000 stills or life pictures and formed a face database called FDID. Besides, we proposed a novel face recognition algorithm and evaluated it with this new database on the real-life videos.

Feature learning is an initial step applied to computer vision tasks and is broadly categorized as: 1) deep feature learning; 2) shallow feature learning. In this paper we focus on shallow feature learning as these algorithms require less computational resources than deep feature learning algorithms. In this paper we propose a shallow feature learning algorithm referred to as Extreme Learning Machine Network (ELMNet). ELMNet is module based neural network consist of feature learning module and a post-processing module. Each feature learning module in ELMNet performs the following operations: 1) patch-based mean removal; 2) ELM auto-encoder (ELM-AE) to learn features. Post-processing module is inserted after the feature learning module and simplifies the features learn by the feature learning modules by hashing and block-wise histogram. Proposed ELMNet outperforms shallow feature learning algorithm PCANet on the MNIST handwritten dataset.

Fully connected multi layer neural networks such as Deep Boltzmann Machines (DBM) performs better than fully connected single layer neural networks in image classification tasks and has a smaller number of hidden layer neurons than Extreme Learning Machine (ELM) based fully connected multi layer neural networks such as Multi Layer ELM (MLELM) and Hierarchical ELM (H-ELM) However, ML-ELM and H-ELM has a smaller training time than DBM. This paper introduces a fully connected multi layer neural network referred to as Multi Layer Multi Objective Extreme Learning Machine (MLMO-ELM) which uses a multi objective formulation to pass the label and non-linear information in order to learn a network model which has a similar number of hidden layer parameters as DBM and smaller training time than DBM. The experimental results show that MLMO-ELM outperforms DBM, ML-ELM and H-ELM on OCR and NORB datasets.

In this paper, we develop a novel algorithm for classifying foreign object debris (FOD) based on the integrated visual features and extreme learning machine (ELM). After image preprocessing, various types of characteristics of the FOD image such as the color names, the scale-invariant feature transform (SIFT), and the histograms of oriented gradient (HOG) features are extracted in the proposed algorithm. These features are then combined into an integrated visual feature vector to characterize foreign objects in the image. Further, according to the extracted integrated visual features, classification is carried out using the ELM classifier. The experimental results show that the proposed classification algorithm outperforms other state-of-the-art methods. Furthermore, we also demonstrate the effectiveness of the use of the integrated visual features and the ELM classifier.

Image distortion can be categorized into two types: content-dependent degradation and content-independent one. Most of the existing perceptual full-reference image quality assessment (IQA) metrics cannot deal with both these two different impacts well. Singular value decomposition (SVD) as a useful mathematical tool that has been used in various image processing applications (e.g., feature extraction). In this paper, SVD is employed to decompose the images into the structural (content-dependent) and the content-independent components. For each portion, a specific assessment model is designed to tailor for its corresponding distortion properties. All the proposed models are then fused to obtain a final quality score by extreme learning machine (ELM), a machine learning technique. Extensive experimental results on six publicly available databases demonstrate that the proposed metric achieves better performance in comparison with the relevant state-of-the-art quality metrics.

Monitoring the presence of occupants in a room in a timely manner is a fundamental step for effective building management. Environmental sensor networks have the advantages of high cost-efficiency and non-intrusiveness on privacy and are very suitable for room occupancy detection. Nonlinear discriminative models, e.g., support vector machine and neural networks, have shown good detection performance due to their ability to model complex relationship. However, they tend to produce unstable detection with frequent fluctuations over time, because they regard training data as independent and ignore the prior knowledge of the room occupancy, i.e., not changing very frequently. To improve the stability of the detection, we propose a two-stage structured learning approach with Extreme Learning Machine (ELM) as the local classifier. In the first stage, ELM is used as a fast nonlinear classifier to obtain preliminary detection results. In the second stage, we form data sequences consisting of the current and previous data points. The preliminary detection results by ELM of the data sequences are then used as input to a linear support vector machine for structured output to generate the final detection results. We test the proposed two-stage structured learning approach on a real-world dataset and show that the proposed approach outperforms the related machine learning methods.

A smile is a common human facial expression used as the indicator for positive emotion. The detection of smiling has many applications, for example, controlling camera shutter when a smile is detected and measuring the degree of satisfaction during a video conference. Many feature extraction methods have been proposed for detecting a smile in the unconstrained scenarios. However, the dimensions of most existing feature descriptors are too huge, which limits their real applications. Moreover, features should be more effective to distinguish between smile face and non-smile face. Motivated by the observation that the mouth shape can effectively reflect a person's smile state, we extracted a novel and snappy set of features that form a feature vector named Pair-wise Distance Vector, which is calculated only based on few points around a mouth. After that, Extreme Learning Machine (ELM) is adopted to classify smile based on these features. The experimental results on GENKI-4K database show that our proposed method outperforms the state-of-the-art methods in terms of accuracy and dimension of features.