Discriminative graph regularized broad learning system for image recognition

Broad learning system (BLS) has been proposed as an alternative method of deep learning. The architecture of BLS is that the input is randomly mapped into series of feature spaces which form the feature nodes, and the output of the feature nodes are expanded broadly to form the enhancement nodes, and then the output weights of the network can be determined analytically. The most advantage of BLS is that it can be learned incrementally without a retraining process when there comes new input data or neural nodes. It has been proven that BLS can overcome the inadequacies caused by training a large number of parameters in gradient-based deep learning algorithms. In this paper, a novel variant graph regularized broad learning system (GBLS) is proposed. Taking account of the locally invariant property of data, which means the similar images may share similar properties, the manifold learning is incorporated into the objective function of the standard BLS. In GBLS, the output weights are constrained to learn more discriminative information, and the classification ability can be further enhanced. Several experiments are carried out to verify that our proposed GBLS model can outperform the standard BLS. What is more, the GBLS also performs better compared with other state-of-the-art image recognition methods in several image databases.     

Discriminative learning and recognition of image set classes using canonical correlations

We address the problem of comparing sets of images for object recognition, where the sets may represent variations in an object's appearance due to changing camera pose and lighting conditions. canonical correlations (also known as principal or canonical angles), which can be thought of as the angles between two d-dimensional subspaces, have recently attracted attention for image set matching. Canonical correlations offer many benefits in accuracy, efficiency, and robustness compared to the two main classical methods: parametric distribution-based and nonparametric sample-based matching of sets. Here, this is first demonstrated experimentally for reasonably sized data sets using existing methods exploiting canonical correlations. Motivated by their proven effectiveness, a novel discriminative learning method over sets is proposed for set classification. Specifically, inspired by classical linear discriminant analysis (LDA), we develop a linear discriminant function that maximizes the canonical correlations of within-class sets and minimizes the canonical correlations of between-class sets. Image sets transformed by the discriminant function are then compared by the canonical correlations. Classical orthogonal subspace method (OSM) is also investigated for the similar purpose and compared with the proposed method. The proposed method is evaluated on various object recognition problems using face image sets with arbitrary motion captured under different illuminations and image sets of 500 general objects taken at different views. The method is also applied to object category recognition using ETH-80 database. The proposed method is shown to outperform the state-of-the-art methods in terms of accuracy and efficiency     

Efficient image signatures and similarities using tensor products of local descriptors

In this paper, we introduce a novel image signature effective in both image retrieval and image classification. Our approach is based on the aggregation of tensor products of discriminant local features, named VLATs (vector of locally aggregated tensors). We also introduce techniques for the packing and the fast comparison of VLATs. We present connections between VLAT and methods like kernel on bags and Fisher vectors. Finally, we show the ability of our method to be effective for two different retrieval problems, thanks to experiments carried out on similarity search and classification datasets.     

Discriminative local binary pattern

Local binary pattern (LBP) is widely used to extract image features as well as motion features in various visual recognition tasks. LBP is formulated in quite a simple form and thus enables us to extract effective features with a low computational cost. There, however, are some limitations mainly regarding sensitivity to noise and loss of image contrast information. In this paper, we propose a novel LBP-based feature extraction method to remedy those drawbacks without degrading the simplicity of the original LBP formulation. LBP is built upon encoding local pixel intensities into binary patterns which can be regarded as separating them into two modes (clusters). We introduce Fisher discriminant criterion to optimize the LBP coding for exploiting binary patterns more stably and discriminatively with robustness to noise. Besides, image contrast information is incorporated in a unified way by leveraging the discriminant score as a weight on the binary pattern; therefore, the prominent patterns, such as around edges, are emphasized. The proposed method is applicable to extract not only image features but also motion features by both efficiently decomposing a XYT volume patch into 2-D patches and employing the effective thresholding strategy based on the volume patch. In the experiments on various visual recognition tasks, the proposed method exhibits superior performance compared to the ordinary LBP and the other methods.     

Discriminative learning can succeed where generative learning fails

Generative algorithms for learning classifiers use training data to separately estimate a probability model for each class. New items are classified by comparing their probabilities under these models. In contrast, discriminative learning algorithms try to find classifiers that perform well on all the training data.We show that there is a learning problem that can be solved by a discriminative learning algorithm, but not by any generative learning algorithm. This statement is formalized using a framework inspired by previous work of Goldberg [P. Goldberg, When can two unsupervised learners achieve PAC separation?, in: Proceedings of the 14th Annual COLT, 2001, pp. 303-319].     

贝叶斯分类器的判别式参数学习

为了提高贝叶斯分类器的分类性能,针对贝叶斯网络分类器的构成特征,提出一种基于参数集成的贝叶斯分类器判别式参数学习算法PEBNC。该算法将贝叶斯分类器的参数学习视为回归问题,将加法回归模型应用于贝叶斯网络分类器的参数学习,实现贝叶斯分类器的判别式参数学习。实验结果表明,在大多数实验数据上,PEBNC能够明显提高贝叶斯分类器的分类准确率。此外,与一般的贝叶斯集成分类器相比,PEBNC不必存储成员分类器的参数,空间复杂度大大降低。     

判别性双向协同表示的图像识别算法

基于协同表示的分类(CRC)以其卓越的协同能力成为人脸分类领域的一个突破。然而在实际应用中,通常只提供很少甚至是单个人脸图像来进行人脸识别,这导致了CRC无法很好地处理光照、表情、姿态和遮挡等问题。针对该问题,提出一种判别性双向协同表示的图像识别算法(DB-CRC)。首先通过引入判别式字典学习(FDDL)模型学习得到一个结构化字典,使得每个特定类的子字典对相关类的样本具有良好的表示能力,由此,较大的类间离散度和较小的类内离散度使得重构误差和编码系数都具有判别性;然后将学习得到的稀疏编码系数作为测试样本数据进行双向表达,建立快速逆向表示模型,利用双向表示策略估计每个测试样本与结构化字典之间的双向重构残差信息;最后利用竞争融合方法对来自双向表示模型的重构残差进行加权排名,实现最终的人脸分类。在AR、PIE、LFW等通用人脸数据库上的实验结果验证了该算法的有效性,特别是对小样本问题的鲁棒性。     

Discriminative learning by sparse representation for classification

Recently, sparsity preserving projections (SPP) algorithm has been proposed, which combines l₁-graph preserving the sparse reconstructive relationship of the data with the classical dimensionality reduction algorithm. However, when applied to classification problem, SPP only focuses on the sparse structure but ignores the label information of samples. To enhance the classification performance, a new algorithm termed discriminative learning by sparse representation projections or DLSP for short is proposed in this paper. DLSP algorithm incorporates the merits of both local interclass geometrical structure and sparsity property. That makes it possess the advantages of the sparse reconstruction, and more importantly, it has better capacity of discrimination, especially when the size of the training set is small. Extensive experimental results on serval publicly available data sets show the feasibility and effectiveness of the proposed algorithm.