Learned Feature Vectors Research Articles

Three-way decisions based feature fusion for Chinese irony detection

Irony detection in social networks has drawn much attention in recent years. In this work, we propose a three-way decisions based feature fusion method for Chinese irony detection in microblog. First, we build a set of discriminating features for Chinese irony detection by considering the characteristics of both Chinese language and social networks, including lexical feature, homophonic pun, consecutive punctuation, length of microblog, passive verbs, and affective imbalance. Second, to tackle the problem of extreme imbalance between different kinds of features, we propose a two-stage classification method for Chinese irony detection by considering feature fusion. In the first stage, a three-way decisions based classifier is trained to predict the ironic tendency based on the lexical feature only. This procedure can reduce the thousands-of-dimensions based lexical feature into a one-dimensional feature. In the second stage, the constructed new feature is merged with other five features into a new feature vector for learning. Experimental results show that our proposed two-stage classification method can get a significant improvement than one-stage classification method.

Hardware Acceleration Implementation of Sparse Coding Algorithm With Spintronic Devices

Sparse coding provides a powerful means to perform feature extraction on high-dimensional data and recently has attracted broad interest for applications. However, it still suffers from a big challenge to realize real-time energy-efficient feature extraction with conventional CPUs/GPUs owing to its highly intensive computation. Benefitting from the non-volatility, low power, high speed, great scalability, and good compatibility with CMOS technology, spintronic devices, such as magnetic tunnel junction and domain wall motion (DWM) device, have been explored for applications, ranging from memory, logic, to neuromorphic computing. In this paper, we explore the possibility of hardware acceleration implementation of the sparse coding algorithm with spintronic devices by a series of design optimizations across the algorithm, architecture, circuit, and device. First, one sparse coding algorithm is selected well suited for parallelization and hardware acceleration. Then, a DWM-based compound spintronic device (CSD) is engineered and modeled, which is envisioned to achieve multiple conductance states. Sequentially, a parallel architecture is presented based on a dense cross-point array of the proposed DWM-based CSD, which can be envisioned to accelerate the selected sparse coding algorithm with a designed dedicated periphery read and write circuit owing to its massively parallel read and write operation. Experimental results show that the selected sparse coding algorithm can be accelerated by 1400× with the proposed parallel architecture in comparison with software implementation. Moreover, its energy dissipation is eight orders of magnitude smaller than that with software implementation. Additionally, an artificial neural circuit (ANC) with the proposed DWM-based CSD is also presented, which can achieve a multi-step transfer function. By using the proposed DWM-based CSD as a single synapse and the proposed ANC as one neuron, a fully forward-connected artificial neural network is constructed to classify the learned feature vectors, and the recognition error rate can reach 7.25%.

Onto2Vec: joint vector-based representation of biological entities and their ontology-based annotations.

MotivationBiological knowledge is widely represented in the form of ontology-based annotations: ontologies describe the phenomena assumed to exist within a domain, and the annotations associate a (kind of) biological entity with a set of phenomena within the domain. The structure and information contained in ontologies and their annotations make them valuable for developing machine learning, data analysis and knowledge extraction algorithms; notably, semantic similarity is widely used to identify relations between biological entities, and ontology-based annotations are frequently used as features in machine learning applications.ResultsWe propose the Onto2Vec method, an approach to learn feature vectors for biological entities based on their annotations to biomedical ontologies. Our method can be applied to a wide range of bioinformatics research problems such as similarity-based prediction of interactions between proteins, classification of interaction types using supervised learning, or clustering. To evaluate Onto2Vec, we use the gene ontology (GO) and jointly produce dense vector representations of proteins, the GO classes to which they are annotated, and the axioms in GO that constrain these classes. First, we demonstrate that Onto2Vec-generated feature vectors can significantly improve prediction of protein–protein interactions in human and yeast. We then illustrate how Onto2Vec representations provide the means for constructing data-driven, trainable semantic similarity measures that can be used to identify particular relations between proteins. Finally, we use an unsupervised clustering approach to identify protein families based on their Enzyme Commission numbers. Our results demonstrate that Onto2Vec can generate high quality feature vectors from biological entities and ontologies. Onto2Vec has the potential to significantly outperform the state-of-the-art in several predictive applications in which ontologies are involved.Availability and implementation https://github.com/bio-ontology-research-group/onto2vec Supplementary information Supplementary data are available at Bioinformatics online.

Entropy and orthogonality based deep discriminative feature learning for object recognition

Inspired by the class-selectivity of the neurons in the inferior temporal (IT) area of the human visual cortex, we propose a novel discriminative feature learning method to improve the object recognition performance of convolutional neural network (CNN) without increasing the network complexity. Specifically, we apply the proposed entropy–orthogonality loss (EOL) to the penultimate layer of the CNN models in the training phase. The EOL explicitly enables the feature vectors learned by a CNN model have the following properties: (1) each dimension of the feature vectors only responds strongly to as few classes as possible, and (2) the feature vectors from different classes are as orthogonal as possible. When combined with the softmax loss, the EOL not only can enlarge the differences in the between-class feature vectors, but also can reduce the variations in the within-class feature vectors. Therefore, the discriminative ability of the learned feature vectors is highly improved. The EOL is general and independent of the CNN structure. Comprehensive experimental comparisons with both the image classification and face verification task on several benchmark datasets demonstrate that utilizing the proposed EOL during training can remarkably improve performance of CNN models compared to the corresponding baseline models trained without utilizing the EOL.

Hashing for Adaptive Real-Time Graph Stream Classification With Concept Drifts.

Many applications involve processing networked streaming data in a timely manner. Graph stream classification aims to learn a classification model from a stream of graphs with only one-pass of data, requiring real-time processing in training and prediction. This is a nontrivial task, as many existing methods require multipass of the graph stream to extract subgraph structures as features for graph classification which does not simultaneously satisfy "one-pass" and "real-time" requirements. In this paper, we propose an adaptive real-time graph stream classification method to address this challenge. We partition the unbounded graph stream data into consecutive graph chunks, each consisting of a fixed number of graphs and delivering a corresponding chunk-level classifier. We employ a random hashing function to compress the original node set of graphs in each chunk for fast feature detection when training chunk-level classifiers. Furthermore, a differential hashing strategy is applied to map unlimited increasing features (i.e., cliques) into a fixed-size feature space which is then used as a feature vector for stochastic learning. Finally, the chunk-level classifiers are weighted in an ensemble learning model for graph classification. The proposed method substantially speeds up the graph feature extraction and avoids unbounded graph feature growth. Moreover, it effectively offsets concept drifts in graph stream classification. Experiments on real-world and synthetic graph streams demonstrate that our method significantly outperforms existing methods in both classification accuracy and learning efficiency.

Robust visual tracking via constrained correlation filter coding

Unconstrained correlation filters based trackers achieve superior performance with high speed in visual tracking. However, such unconstrained correlation filters do not impose any hard constraint to their responses to have a certain value, which brings about classification ambiguity on intractable samples (i.e., two similar samples from different classes). To tackle this issue, in this paper, constrained correlation filter is introduced into visual tracking framework to alleviate classification ambiguity for more accurate target location. By imposing distinguishable hard constraints on the response map to different classes, a supervised coding method is proposed to encode various candidate samples by a discriminative filter bank. The learned high-level feature vectors are sent to a Naive Bayes classifier to separate the target from the background. Besides, parameters updating schemes in the constrained filter and classifier are introduced to adapt to appearance changes of the target with less possibility of drifting. Both qualitative and quantitative evaluations on Object Tracking Benchmark (OTB) show that the proposed tracking method achieves favorable performance compared with other state-of-the-art methods.

A topological approach for protein classification

Abstract Protein function and dynamics are closely related to its sequence and structure.However, prediction of protein function and dynamics from its sequence and structure is still a fundamental challenge in molecular biology. Protein classification, which is typically done through measuring the similarity between proteins based on protein sequence or physical information, serves as a crucial step toward the understanding of protein function and dynamics. Persistent homology is a new branch of algebraic topology that has found its success in the topological data analysis in a variety of disciplines, including molecular biology. The present work explores the potential of using persistent homology as an independent tool for protein classification. To this end, we propose a molecular topological fingerprint based support vector machine (MTF-SVM) classifier. Specifically,we construct machine learning feature vectors solely fromprotein topological fingerprints,which are topological invariants generated during the filtration process. To validate the presentMTF-SVMapproach, we consider four types of problems. First, we study protein-drug binding by using the M2 channel protein of influenza A virus. We achieve 96% accuracy in discriminating drug bound and unbound M2 channels. Secondly, we examine the use of MTF-SVM for the classification of hemoglobin molecules in their relaxed and taut forms and obtain about 80% accuracy. Thirdly, the identification of all alpha, all beta, and alpha-beta protein domains is carried out using 900 proteins.We have found a 85% success in this identification. Finally, we apply the present technique to 55 classification tasks of protein superfamilies over 1357 samples and 246 tasks over 11944 samples. Average accuracies of 82% and 73% are attained. The present study establishes computational topology as an independent and effective alternative for protein classification.

Coupled Discriminative Feature Learning for Heterogeneous Face Recognition

This paper presents a coupled discriminative feature learning (CDFL) method for heterogeneous face recognition (HFR). Different from most existing HFR approaches which use hand-crafted feature descriptors for face representation, our CDFL directly learns discriminative features from raw pixels for face representation. In particular, a couple of image filters is learned in CDFL to simultaneously exploit discriminative information and to reduce the appearance difference of face images captured across different modalities. With the help of the learned filters, CDFL can maximize the interclass variations and minimize the intraclass variations of the learned feature vectors, and meanwhile maximize the correlation of face images of the same person from different modalities by solving a generalized eigenvalue problem. Experimental results on three different heterogeneous face recognition applications show the effectiveness of our proposed approach.

Scene classification based on single-layer SAE and SVM

Scene classification aims to group images into semantic categories. It is a challenging problem in computer vision due to the difficulties of intra-class variability and inter-class similarity. In this paper, a scene classification approach based on single-layer sparse autoencoder (SAE) and support vector machine (SVM) is proposed. This approach consists of two steps: SAE-based feature learning step and SVM-based classification step. In the first step, a single-layer SAE network is constructed and trained by the patches which are sampled randomly from the source images. The feature representation of images is learned by the trained single-layer SAE network. Meanwhile, a pooling operation is used to reduce the dimension of the learned feature vectors. In the second step, in order to improve the classification accuracy, the parameters of SVM are optimized by a particle swarm optimization (PSO) based algorithm. The one-versus-one strategy is employed for the multiple scene classification problem. To show the efficiency of the proposed approach, several public data sets are employed. The results reveal that the proposed approach achieves better classification accuracy than the existing state-of-the-art methods.

Non-negative matrix factorization based methods for object recognition

Non-negative matrix factorization (NMF) is a new feature extraction method. But the learned feature vectors are not directly suitable for further analysis such as object recognition using the nearest neighbor classifier in contrast to traditional principal component analysis (PCA) because the learned bases are not orthonormal to each other. This paper investigates how to improve the accuracy of recognition based on this new method from two viewpoints. One is to adopt a Riemannian metric like distance for the learned feature vectors instead of Euclidean distance. The other is to first orthonormalize the learned bases and then to use the projections of data based on the orthonormalized bases for further recognition. Experiments on the USPS database demonstrate the proposed methods can improve accuracy and even outperform PCA. We believe that the proposed methods can make NMF used as widely as PCA.