Taxonomy-Augmented Features for Document Clustering

Feature selection techniques have become an apparent need for diagnostic methods such as a least squares support vector machine (LS-SVM). Most researchers use wavelet transform coefficients of the time-domain transient response data obtained from filtered analog circuits as features to train a LS-SVM classifier to diagnose faults. But wavelet coefficient features have certain disadvantages such as no physical meanings. Thus, in this paper, two new feature vectors with clearly defined meanings based on a time-domain response curve and a frequency response curve of a filter are proposed, respectively. In addition, a statistical property feature vector which represents global properties of the time-domain response curve or the frequency response curve is proposed. The results from the simulation data and real data for a biquad filter showed the following: (1) these proposed conventional time-domain and frequency features, which are already familiar to designers of filtered analog circuits, have good diagnostic accuracy—all above 91% for the example circuit; (2) the best accuracies using the proposed statistical property feature vector are 100% for time-domain simulation data, and for both real experiment data ; (3) the diagnostic accuracy using the proposed combined feature vector is more accurate than conventional feature vectors; (4) an LS-SVM can be used to diagnose faults in a real analog circuit that only has a few fault samples.

An effective and widespread approach for shape-based 3D model retrieval (3DMR) is to use a feature vector per 3D model obtained by aggregating, or pooling, a set of local features extracted from the 3D model. State-of-the-art feature aggregation algorithms, such as Fisher Vector (FV) coding [7] or Super Vector (SV) coding [22], used in the approach is not spatially efficient, however. The FV or SV, for example, typically encodes a local feature into a very high-dimensional (e.g., 300k-dimensional) vector. For a database containing a large number of 3D models, the spatial cost of storing all the aggregated feature vectors for the database becomes very high. In this paper, we propose a novel, spatially efficient yet accurate feature aggregation algorithm called Sum of Sparse Binary codes (SSB) aggregation. The SSB first encodes a local feature into a highly sparse binary code. Then, a set of sparse binary codes are aggregated efficiently by simple summing into a compact feature vector. We also propose fast SSB (fSSB) aggregation, which is a computationally efficient approximation of the SSB. Experiments using a 3DMR scenario show that the proposed algorithms are significantly more efficient than the state-of-the-art feature aggregation algorithms we have compared against. At the same time, retrieval accuracies of the proposed algorithms are equal or better than the state-of-the-art aggregation algorithms.

We have developed a new technique, which may enable a more accurate, complete characterization of the circulatory system, including local or global hydrodynamic phenomena, using multiple measurements from several anatomic locations and/or multiple modalities. This technique, Laguerre model blind system identification (LMBSI), uses a Laguerre function series expansion to provide a compact but complete quantitative description of the distinct behavior of two or more circulatory waveforms. LMBSI identifies a set of five parameters per channel plus one common parameter that can be treated as a feature vector and used to predict cardiovascular parameters of interest. Standard statistical techniques can be used to extract information from that compact feature vector. In this paper, multiparameter regression is used to predict cardiac output, using two separate arterial pressure waveforms and the LMBSI algorithm. This serves as a proof-of-principle that two distinct circulatory waveforms, with LMBSI, can be used to characterize the circulatory system. In the future, this technique might be applied to noninvasive circulatory measurements.

We present in this paper some lightweight metrics for quality assessment of fingerprint samples captured from a general-purposed optical camera using block-based autocorrelation and fast Walsh-Hadamard spectrum features. We generate a feature vector including 6 components in three categories to assess an image block's quality: (1) gray-scale values statistics, (2) autocorrelation based features, (3) Walsh-Hadamard spectrum features from the autocorrelation result. Then a support vector machine is adopted to produce a binary decision whether the image block should be classified as high-quality fingerprint area or not. Finally our approach generates a quality score by counting the high-quality blocks in the sample image. Compared to our previous work [12] on the same topic, this paper uses a more compact feature vector and less computationally-complicated transformation. Experiments demonstrated the proposed metrics' effectiveness.

As the scale of 3D model databases increase, speed, in addition to accuracy, of its search becomes very important. One way to achieve fast search is to use a compact 3D shape feature whose cost of comparison is very small. Binarization of a real-valued feature vector, via hashing, is a way to obtain such compact feature vector. Previous algorithms for producing binarized 3D model features via hashing consisted of two disconnected stages; handcrafted real-valued 3D shape feature extraction followed by hashing into binary code. This compartmentalized approach, however, leads to less-than optimal binary codes, as these two stages are optimized independently. This paper proposes a deep semantic hashing algorithm called Binarized Deep Local feature Aggregation Network (BDLAN) which jointly optimizes real-valued feature extraction per 3D model and its banarization via hashing. BDLAN training minimizes quantization error caused by binarization. However, this constraint alone often maps real-valued features to their nearest binary codes in Hamming space, which are nonoptimal local minima. To alleviate the issue, we add a simple regularization called Probabilistic Bit Inversion (PBI) of binary codes. Experimental evaluation of the proposed algorithms demonstrates superior efficiency and competitive accuracy to the existing 3D model retrieval algorithms employing real-valued features.

Accurately learning from user data while providing quantifiable privacy guarantees provides an opportunity to build better ML models while maintaining user trust. This paper presents a formal approach to carrying out privacy preserving text perturbation using the notion of d_χ-privacy designed to achieve geo-indistinguishability in location data. Our approach applies carefully calibrated noise to vector representation of words in a high dimension space as defined by word embedding models. We present a privacy proof that satisfies d_χ-privacy where the privacy parameter $\varepsilon$ provides guarantees with respect to a distance metric defined by the word embedding space. We demonstrate how $\varepsilon$ can be selected by analyzing plausible deniability statistics backed up by large scale analysis on GloVe and fastText embeddings. We conduct privacy audit experiments against $2$ baseline models and utility experiments on 3 datasets to demonstrate the tradeoff between privacy and utility for varying values of varepsilon on different task types. Our results demonstrate practical utility (

Local binary patterns (LBP) represent a well-established technique for reliable facial recognition. Multi-scale block (MB) LBP, which process average pixel values of block sub-regions instead of single pixels, have been proposed in order to achieve more robust feature vectors, which consist of histograms of (MB)-LBP values. In this work, we present a simple, yet effective, method to binarize histograms obtained from a MB-LBP face recognition system. The proposed method extracts compact binary feature vectors which allow for a rapid comparison and which are required for privacy enhanced biometric processing, when biometric template protection is enabled. On the publicly available FERET and Extended-Yale-B dataset, the presented scheme reveals only a negligible drop in biometric performance compared to the original MB-LBP system, while a substantial speed-up is achieved in the comparison stage.

The information fusion of face and palmprint biometrics using local features is investigated at feature level. The proposed method uses local information extracted from local region of biometric image which has rich statistical information. The texture of each region is processed using multiresolution analysis with different orientations and scales. The feature dimensionality of each region is reduced to produce a compact and high discriminative feature vector used for concatenation process. Feature fusion of the extracted features is able to increase the discrimination power in the feature space. The use of principle component analysis (PCA) and linear discriminant analysis (LDA) methods significantly reduce dimension of the feature vector by removing redundant and noise data while increasing the discriminant power in the fused feature space. Results of both identification and verification rates show significant improvement compared to that achieved by single modal biometrics and several existing multimodal methods.

Word vectors are at the core of many natural language processing tasks. Recently, there has been interest in post-processing word vectors to enrich their semantic information. In this paper, we introduce a novel word vector post-processing technique based on matrix conceptors (Jaeger 2014), a family of regularized identity maps. More concretely, we propose to use conceptors to suppress those latent features of word vectors having high variances. The proposed method is purely unsupervised: it does not rely on any corpus or external linguistic database. We evaluate the post-processed word vectors on a battery of intrinsic lexical evaluation tasks, showing that the proposed method consistently outperforms existing state-of-the-art alternatives. We also show that post-processed word vectors can be used for the downstream natural language processing task of dialogue state tracking, yielding improved results in different dialogue domains.

Recently, Content-based medical image retrieval (CBMIR) systems enable fast diagnosis via the assessment of the visual information in medical application. Most of the state-of-the-art CBMIR systems facing few issues: computationally expensive due to the usage of high dimensional feature vectors and complex classifier/clustering approaches. The reasons behind this are, inability to properly handle the “semantic gap” and the high intra-class versus inter-class variability problem of the medical image database (like radiographic image database). This yields a crucial demand for developing computationally efficient and highly effective retrieval system. For this purpose, the present study proposed an efficient retrieval system which has a four-fold: First, pre-processing and Feature extraction of input image using canonical correlation analysis (CCA). By this approach extracted the feature in both pixel and feature domains and examined more rigorously. Second, applied Fuzzy C means clustering of pixel intensity values as features based on the singular value decomposition. Through this can grouping, the image based on the pixel intensity value. Third, deep convolutional neural network with SVM classifier which makes implementable and requires only a compact feature vector representation of the stored database image with their class levels during retrieval. Finally evaluated the performance based on the measure of Mean Average Precision, Correct rate (CR), Error rate (ER), Accuracy. The classification results and learned features are used for the purpose of retrieving the medical images in a database. The proposed retrieval system performs better than the traditional approach in terms of measuring average value of precision, recall, f-measure and accuracy 95.9%, 94.96%, 95.37% and 95.798% respectively. The suggested approach is best suited towards retrieving the medical images for various part of the body.

The written notes in the Electronic Health Records contain a vast amount of information about patients. Implementing automated approaches for text classification tasks requires the automated methods to be well-interpretable, and topic models can be used for this goal as they can indicate what topics in a text are relevant to making a decision. We propose a new topic modeling algorithm, FLSA-E, and compare it with another state-of-the-art algorithm FLSA-W. In FLSA-E, topics are found by fuzzy clustering in a word embedding space. Since we use word embeddings as the basis for our clustering, we extend our evaluation with word-embeddings-based evaluation metrics. We find that different evaluation metrics favour different algorithms. Based on the results, there is evidence that FLSA-E has fewer outliers in its topics, a desirable property, given that within-topic words need to be semantically related.

Question classification in Persian using word vectors and frequencies

In this paper, an algorithm based on Bi-directional Gated Recurrent Unit (Bi-GRU) and Gradient Boosting Decision Tree (GBDT) is proposed to extract the entity relationship of Traditional Chinese Medicine influenza. Firstly, the word vector is used as the input data set and the word vector is constructed by the word embedding model Word2Vec tool. Then the sentence feature is extracted by Bi-GRU and the attention mechanism is integrated to improve the accuracy of feature extraction. Finally, the feature vector is input into the GBDT algorithm for classification training and prediction to complete the Traditional Chinese Medicine influenza entity relationship extraction. In this paper, a variety of different entity relation extraction algorithms are compared with this algorithm to verify the effectiveness of the algorithm. This algorithm improves the stability of the model and effectively solves the problem of insufficient generalization ability of the model. Therefore, when studying the relationship extraction of Traditional Chinese Medicine texts, we can give priority to using Bi-GRU+GBDT model. Also through the experiment to adjust the model parameters and comparison to get the optimal parameters of Traditional Chinese Medicine influenza relationship extraction experiment.

In this paper, we conduct in‐depth research and analysis on the intelligent recognition and teaching of English fuzzy text through parallel projection and region expansion. Multisense Soft Cluster Vector (MSCVec), a multisense word vector model based on nonnegative matrix decomposition and sparse soft clustering, is constructed. The MSCVec model is a monolingual word vector model, which uses nonnegative matrix decomposition of positive point mutual information between words and contexts to extract low‐rank expressions of mixed semantics of multisense words and then uses sparse. It uses the nonnegative matrix decomposition of the positive pointwise mutual information between words and contexts to extract the low‐rank expressions of the mixed semantics of the polysemous words and then uses the sparse soft clustering algorithm to partition the multiple word senses of the polysemous words and also obtains the global sense of the polysemous word affiliation distribution; the specific polysemous word cluster classes are determined based on the negative mean log‐likelihood of the global affiliation between the contextual semantics and the polysemous words, and finally, the polysemous word vectors are learned using the Fast text model under the extended dictionary word set. The advantage of the MSCVec model is that it is an unsupervised learning process without any knowledge base, and the substring representation in the model ensures the generation of unregistered word vectors; in addition, the global affiliation of the MSCVec model can also expect polysemantic word vectors to single word vectors. Compared with the traditional static word vectors, MSCVec shows excellent results in both word similarity and downstream text classification task experiments. The two sets of features are then fused and extended into new semantic features, and similarity classification experiments and stack generalization experiments are designed for comparison. In the cross‐lingual sentence‐level similarity detection task, SCLVec cross‐lingual word vector lexical‐level features outperform MSCVec multisense word vector features as the input embedding layer; deep semantic sentence‐level features trained by twin recurrent neural networks outperform the semantic features of twin convolutional neural networks; extensions of traditional statistical features can effectively improve cross‐lingual similarity detection performance, especially cross‐lingual topic model (BL‐LDA); the stack generalization integration approach maximizes the error rate of the underlying classifier and improves the detection accuracy.

Neural network models have been demonstrated to be capable of achieving state-of-the-art performance in sentence sentiment classification. Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs) are two widely used neural network models for NLP. However, since sentences consist of the same words in different order may represent different meaning in sentiment, it cannot be neglected that the word embedding training model ignores the factor of word order in sentence to quicken the training process. In this work, we mainly consider that word order is important for sentence sentiment classification, designing an encode-decode model called CNN-LSTM combined the strength of CNN and LSTM to demonstrate that word order of sentence plays an important role in sentiment analysis based on the word embedding which is designed as an order_w2v model taking in word order during word2vec training process. We evaluate the CNN-LSTM and order_w2v in sentiment classification both on Chinese and English datasets. The experimental results verify that the model considering the word order can achieve better results in sentiment analysis.