Consider a classification problem where we have both labeled and unlabeled data available. We show that for linear classifiers defined by convex margin-based surrogate losses that are decreasing, it is impossible to construct any semi-supervised approach that is able to guarantee an improvement over the supervised classifier measured by this surrogate loss on the labeled and unlabeled data. For convex margin-based loss functions that also increase, we demonstrate safe improvements are possible.
In this paper, a model for classification of logos based on symbolic representation of features is presented. The proposed model makes use of global features of logo images such as color, texture, and shape features for classification. The logo images are broadly classified into three different classes, viz., logo image containing only text, an image with only symbol, and an image with both text and a symbol. In each class, the similar looking logo images are clustered using K-means clustering algorithm. The intra-cluster variations present in each cluster corresponding to each class are then preserved using symbolic interval data. Thus referenced logo images are represented in the form of interval data. A sample logo image is then classified using suitable symbolic classifier. For experimentation purpose, relatively large amount of color logo images is created consisting of 5044 logo images. The classification results are validated with the help of accuracy, precision, recall, F-measure, and time. To check the efficacy of the proposed model, the comparative analyses are given against the other models. The results show that the proposed model outperforms the other models with respect to time and F-measure.
Classification of human behavior is key to developing closed-loop Deep Brain Stimulation (DBS) systems, which may be able to decrease the power consumption and side effects of the existing systems. Recent studies have shown that the Local Field Potential (LFP) signals from both Subthalamic Nuclei (STN) of the brain can be used to recognize human behavior. Since the DBS leads implanted in each STN can collect three bipolar signals, the selection of a suitable pair of LFPs that achieves optimal recognition performance is still an open problem to address. Considering the presence of synchronized aggregate activity in the basal ganglia, this paper presents an FFT-based synchronization approach to automatically select a relevant pair of LFPs and use the pair together with an SVM-based MKL classifier for behavior recognition purposes. Our experiments on five subjects show the superiority of the proposed approach compared to other methods used for behavior classification.
Gene expression data is widely used in disease analysis and cancer diagnosis. However, since gene expression data could contain thousands of genes simultaneously, successful microarray classification is rather difficult. Feature selection is an important pre-treatment for any classification process. Selecting a useful gene subset as a classifier not only decreases the computational time and cost, but also increases classification accuracy. In this study, we applied the information gain method as a filter approach, and an improved binary particle swarm optimization as a wrapper approach to implement feature selection; selected gene subsets were used to evaluate the performance of classification. Experimental results show that by employing the proposed method fewer gene subsets needed to be selected and better classification accuracy could be obtained.
A brain-computer interface (BCI) is a system that aims for establishing a non-muscular communication path for subjects who had suffer from a neurodegenerative disease. Many BCI systems make use of the phenomena of event-related synchronization and de-synchronization of brain waves as a main feature for classification of different cognitive tasks. However, the temporal dynamics of the electroencephalographic (EEG) signals contain additional information that can be incorporated into the inference engine in order to improve the performance of the BCIs. This information about the dynamics of the signals have been exploited previously in BCIs by means of generative and discriminative methods. In particular, hidden Markov models (HMMs) have been used in previous works. These methods have the disadvantage that the model parameters such as the number of hidden states and the number of Gaussian mixtures need to be fix "a priori". In this work, we propose a Bayesian nonparametric model for brain signal classification that does not require "a priori" selection of the number of hidden states and the number of Gaussian mixtures of a HMM. The results show that the proposed model outperform other methods based on HMM as well as the winner algorithm of the BCI competition IV.
In this paper, we extend some usual techniques of classification resulting from a large-scale data-mining and network approach. This new technology, which in particular is designed to be suitable to big data, is used to construct an open consolidated database from raw data on 4 million patents taken from the US patent office from 1976 onward. To build the pattern network, not only do we look at each patent title, but we also examine their full abstract and extract the relevant keywords accordingly. We refer to this classification as semantic approach in contrast with the more common technological approach which consists in taking the topology when considering US Patent office technological classes. Moreover, we document that both approaches have highly different topological measures and strong statistical evidence that they feature a different model. This suggests that our method is a useful tool to extract endogenous information.
This paper presents a deep nonlinear metric learning framework for data visualization on an image dataset. We propose the Triangular Similarity and prove its equivalence to the Cosine Similarity in measuring a data pair. Based on this novel similarity, a geometrically motivated loss function - the triangular loss - is then developed for optimizing a metric learning system comprising two identical CNNs. It is shown that this deep nonlinear system can be efficiently trained by a hybrid algorithm based on the conventional backpropagation algorithm. More interestingly, benefiting from classical manifold learning theories, the proposed system offers two different views to visualize the outputs, the second of which provides better classification results than the state-of-the-art methods in the visualizable spaces.
It has long been recognized that the invariance and equivariance properties of a representation are critically important for success in many vision tasks. In this paper we present Steerable Convolutional Neural Networks, an efficient and flexible class of equivariant convolutional networks. We show that steerable CNNs achieve state of the art results on the CIFAR image classification benchmark. The mathematical theory of steerable representations reveals a type system in which any steerable representation is a composition of elementary feature types, each one associated with a particular kind of symmetry. We show how the parameter cost of a steerable filter bank depends on the types of the input and output features, and show how to use this knowledge to construct CNNs that utilize parameters effectively.
Nowadays the CNN is widely used in practical applications for image classification task. However the design of the CNN model is very professional work and which is very difficult for ordinary users. Besides, even for experts of CNN, to select an optimal model for specific task may still need a lot of time (to train many different models). In order to solve this problem, we proposed an automated CNN recommendation system for image classification task. Our system is able to evaluate the complexity of the classification task and the classification ability of the CNN model precisely. By using the evaluation results, the system can recommend the optimal CNN model and which can match the task perfectly. The recommendation process of the system is very fast since we don't need any model training. The experiment results proved that the evaluation methods are very accurate and reliable.
We introduce YOLO9000, a state-of-the-art, real-time object detection system that can detect over 9000 object categories. First we propose various improvements to the YOLO detection method, both novel and drawn from prior work. The improved model, YOLOv2, is state-of-the-art on standard detection tasks like PASCAL VOC and COCO. At 67 FPS, YOLOv2 gets 76.8 mAP on VOC 2007. At 40 FPS, YOLOv2 gets 78.6 mAP, outperforming state-of-the-art methods like Faster RCNN with ResNet and SSD while still running significantly faster. Finally we propose a method to jointly train on object detection and classification. Using this method we train YOLO9000 simultaneously on the COCO detection dataset and the ImageNet classification dataset. Our joint training allows YOLO9000 to predict detections for object classes that don't have labelled detection data. We validate our approach on the ImageNet detection task. YOLO9000 gets 19.7 mAP on the ImageNet detection validation set despite only having detection data for 44 of the 200 classes. On the 156 classes not in COCO, YOLO9000 gets 16.0 mAP. But YOLO can detect more than just 200 classes; it predicts detections for more than 9000 different object categories. And it still runs in real-time.
Automatic profiling of social media users is an important task for supporting a multitude of downstream applications. While a number of studies have used social media content to extract and study collective social attributes, there is a lack of substantial research that addresses the detection of a user's industry. We frame this task as classification using both feature engineering and ensemble learning. Our industry-detection system uses both posted content and profile information to detect a user's industry with 64.3% accuracy, significantly outperforming the majority baseline in a taxonomy of fourteen industry classes. Our qualitative analysis suggests that a person's industry not only affects the words used and their perceived meanings, but also the number and type of emotions being expressed.
Obtaining common representations from different modalities is important in that they are interchangeable with each other in a classification problem. For example, we can train a classifier on image features in the common representations and apply it to the testing of the text features in the representations. Existing multi-modal representation learning methods mainly aim to extract rich information from paired samples and train a classifier by the corresponding labels; however, collecting paired samples and their labels simultaneously involves high labor costs. Addressing paired modal samples without their labels and single modal data with their labels independently is much easier than addressing labeled multi-modal data. To obtain the common representations under such a situation, we propose to make the distributions over different modalities similar in the learned representations, namely modality-invariant representations. In particular, we propose a novel algorithm for modality-invariant representation learning, named Deep Modality Invariant Adversarial Network (DeMIAN), which utilizes the idea of Domain Adaptation (DA). Using the modality-invariant representations learned by DeMIAN, we achieved better classification accuracy than with the state-of-the-art methods, especially for some benchmark datasets of zero-shot learning.
Most of the temporal lobe epilepsy detection approaches are based on hippocampus deformation and use complicated features, resulting, detection is done with complicated features extraction and pre-processing task. In this paper, a new detection method based on shape-based features and spherical harmonics is proposed which can analysis the hippocampus shape anomaly and detection asymmetry. This method consisted of two main parts; (1) shape feature extraction, and (2) image classification. For evaluation, HFH database is used which is publicly available in this field. Nine different geometry and 256 spherical harmonic features are introduced then selected Eighteen of them that detect the asymmetry in hippocampus significantly in a randomly selected subset of the dataset. Then a support vector machine (SVM) classifier was employed to classify the remaining images of the dataset to normal and epileptic images using our selected features. On a dataset of 25 images, 12 images were used for feature extraction and the rest 13 for classification. The results show that the proposed method has accuracy, specificity and sensitivity of, respectively, 84%, 100%, and 80%. Therefore, the proposed approach shows acceptable result and is straightforward also; complicated pre-processing steps were omitted compared to other methods.
The standard approach to supervised classification involves the minimization of a log-loss as an upper bound to the classification error. While this is a tight bound early on in the optimization, it overemphasizes the influence of incorrectly classified examples far from the decision boundary. Updating the upper bound during the optimization leads to improved classification rates while transforming the learning into a sequence of minimization problems. In addition, in the context where the classifier is part of a larger system, this modification makes it possible to link the performance of the classifier to that of the whole system, allowing the seamless introduction of external constraints.
Single modality action recognition on RGB or depth sequences has been extensively explored recently. It is generally accepted that each of these two modalities has different strengths and limitations for the task of action recognition. Therefore, analysis of the RGB+D videos can help us to better study the complementary properties of these two types of modalities and achieve higher levels of performance. In this paper, we propose a new deep autoencoder based shared-specific feature factorization network to separate input multimodal signals into a hierarchy of components. Further, based on the structure of the features, a structured sparsity learning machine is proposed which utilizes mixed norms to apply regularization within components and group selection between them for better classification performance. Our experimental results show the effectiveness of our cross-modality feature analysis framework by achieving state-of-the-art accuracy for action classification on five challenging benchmark datasets.
The monitoring of sleep patterns without patient's inconvenience or involvement of a medical specialist is a clinical question of significant importance. To this end, we propose an automatic sleep stage monitoring system based on an affordable, unobtrusive, discreet, and long-term wearable in-ear sensor for recording the Electroencephalogram (ear-EEG). The selected features for sleep pattern classification from a single ear-EEG channel include the spectral edge frequency (SEF) and multi- scale fuzzy entropy (MSFE), a structural complexity feature. In this preliminary study, the manually scored hypnograms from simultaneous scalp-EEG and ear-EEG recordings of four subjects are used as labels for two analysis scenarios: 1) classification of ear-EEG hypnogram labels from ear-EEG recordings and 2) prediction of scalp-EEG hypnogram labels from ear-EEG recordings. We consider both 2-class and 4-class sleep scoring, with the achieved accuracies ranging from 78.5 % to 95.2 % for ear-EEG labels predicted from ear-EEG, and 76.8 % to 91.8 % for scalp-EEG labels predicted from ear-EEG. The corresponding kappa coefficients, which range from 0.64 to 0.83 for Scenario 1 and from 0.65 to 0.80 for Scenario 2, indicate a Substantial to Almost Perfect agreement, thus proving the feasibility of in-ear sensing for sleep monitoring in the community.
Today, smartphone devices are owned by a large portion of the population and have become a very popular platform for accessing the Internet. Smartphones provide the user with immediate access to information and services. However, they can easily expose the user to many privacy risks. Applications that are installed on the device and entities with access to the device's Internet traffic can reveal private information about the smartphone user and steal sensitive content stored on the device or transmitted by the device over the Internet. In this paper, we present a method to reveal various demographics and technical computer skills of smartphone users by their Internet traffic records, using machine learning classification models. We implement and evaluate the method on real life data of smartphone users and show that smartphone users can be classified by their gender, smoking habits, software programming experience, and other characteristics.
With the rise in militant activity and rogue behaviour in oil and gas regions around the world, oil pipeline disturbances is on the increase leading to huge losses to multinational operators and the countries where such facilities exist. However, this situation can be averted if adequate predictive monitoring schemes are put in place. We propose in the first part of this paper, an artificial intelligence predictive monitoring system capable of predictive classification and pattern recognition of pipeline datasets. The predictive system is based on a highly sparse predictive Deviant Learning Algorithm (p-DLA) designed to synthesize a sequence of memory predictive clusters for eventual monitoring, control and decision making. The DLA (p-DLA) is compared with a popular machine learning algorithm, the Long Short-Term Memory (LSTM) which is based on a temporal version of the standard feed-forward back-propagation trained artificial neural networks (ANNs). The results of simulations study show impressive results and validates the sparse memory predictive approach which favours the sub-synthesis of a highly compressed and low dimensional knowledge discovery and information prediction scheme. It also shows that the proposed new approach is competitive with a well-known and proven AI approach such as the LSTM.
In this paper we introduce a fuzzy constraint linear discriminant analysis (FC-LDA). The FC-LDA tries to minimize misclassification error based on modified perceptron criterion that benefits handling the uncertainty near the decision boundary by means of a fuzzy linear programming approach with fuzzy resources. The method proposed has low computational complexity because of its linear characteristics and the ability to deal with noisy data with different degrees of tolerance. Obtained results verify the success of the algorithm when dealing with different problems. Comparing FC-LDA and LDA shows superiority in classification task.
This paper describes a number of fundamental and practical problems in the application of hidden-Markov models and Bayes when applied to cursive-script recognition. Several problems, however, will have an effect in other application areas. The most fundamental problem is the propagation of error in the product of probabilities. This is a common and pervasive problem which deserves more attention. On the basis of Monte Carlo modeling, tables for the expected relative error are given. It seems that it is distributed according to a continuous Poisson distribution over log probabilities. A second essential problem is related to the appropriateness of the Markov assumption. Basic tests will reveal whether a problem requires modeling of the stochastics of seriality, at all. Examples are given of lexical encodings which cover 95-99% classification accuracy of a lexicon, with removed sequence information, for several European languages. Finally, a summary of results on a non- Bayes, non-Markov method in handwriting recognition are presented, with very acceptable results and minimal modeling or training requirements using nearest-mean classification.