Self-Attention Networks and Adaptive Support Vector Machine for Aspect-Level Sentiment Classification

Aspect-level sentiment classification aims to integrating the context to predict the sentiment polarity of aspect-specific in a text, which has been quite useful and popular, e.g. opinion survey and products’ recommending in e-commerce. Many recent studies exploit a Long Short-Term Memory (LSTM) networks to perform aspect-level sentiment classification, but the limitation of long-term dependencies is not solved well, so that the semantic correlations between each two words of the text are ignored. In addition, traditional classification model adopts SoftMax function based on probability statistics as classifier, but ignores the words’ features in the semantic space. Support Vector Machine (SVM) can fully use the information of characteristics and it is appropriate to make classification in the high dimension space, however which just considers the maximum distance between different classes and ignores the similarities between different features of the same classes. To address these defects, we propose the two-stages novel architecture named Self Attention Networks and Adaptive SVM (SAN-ASVM) for aspect-level sentiment classification. In the first-stage, in order to overcome the long-term dependencies, Multi-Heads Self Attention (MHSA) mechanism is applied to extract the semantic relationships between each two words, furthermore 1-hop attention mechanism is designed to pay more attention on some important words related to aspect-specific. In the second-stage, ASVM is designed to substitute the SoftMax function to perform sentiment classification, which can effectively make multi-classifications in high dimensional space. Extensive experiments on SemEval2014, SemEval2016 and Twitter datasets are conducted, compared experiments prove that SAN-ASVM model can obtains better performance.

Increased persuadability and credulity in people with corpus callosum dysgenesis

Corpus callosum dysgenesis is one of the most common congenital neurological malformations. Despite being a clear and identifiable structural alteration of the brains white matter connectivity, the impact of corpus callosum dysgenesis on cognition and behavior has remained unclear. Here we build upon past clinical observations in the literature to define the clinical phenotype of corpus callosum dysgenesis better using unadjusted and adjusted group differences compared with a neurotypical sample on a range of social and cognitive measures that have been previously reported to be impacted by a corpus callosum dysgenesis diagnosis. Those with a diagnosis of corpus callosum dysgenesis (n = 22) demonstrated significantly higher persuadability, credulity, and insensitivity to social trickery than neurotypical (n = 86) participants, after controlling for age, sex, education, autistic-like traits, social intelligence, and general cognition. To explore this further, machine learning, utilizing a set neurotypical sample for training the normative covariance structure of our psychometric variables, was used to test whether these dimensions possessed the capability to discriminate between a test-set of neurotypical and corpus callosum dysgenesis participants. We found that participants with a diagnosis of corpus callosum dysgenesis were best classed within dimension space along the same axis as persuadability, credulity, and insensitivity to social trickery after controlling for age and sex, with Leave-One-Out-Cross-Validation across 250 training-set permutations providing a mean accuracy of 71.7 percent. These results have wide-reaching implications for a) the characterization of corpus callosum dysgenesis, and b) the role of the corpus callosum in social inference.

A note on the high-dimensional sparse Fourier transform in the continuous setting

In this paper, we theoretically propose a new hashing scheme to establish the sparse Fourier transform in high-dimension space. The estimation of the algorithm complexity shows that this sparse Fourier transform can overcome the curse of dimensionality. To the best of our knowledge, this is the first polynomial-time algorithm to recover the high-dimensional continuous frequencies.

Optimized Polynomial Classifier for Classification of M-PSK Signals

Automatic modulation classification (AMC) is the emerging research area for military and civil applications. In this paper, M-PSK signals are classified using the optimized polynomial classifier. The distinct features i.e., higher order cumulants (HOC’s) are extracted from the noisy received signal and the dataset is generated with different number of samples, various SNR’s and on several fading channels. The proposed classifier structure classifies the overall modulation classification problem into binary sub-classifications. In each sub-classification, the extracted features are expanded using polynomial expansion into higher dimension space. In higher dimension space numerous non-linearly separable classes becomes linearly separable. The performance of the proposed classifier is evaluated on Rayleigh and Rician fading channels in the presence of additive white gaussian noise (AWGN). The polynomial classifier performance is optimized using one of the famous heuristic computational techniques i.e., Genetic Algorithm (GA). The extensive simulations have been carried with and without optimization, which shows relatively better percentage classification accuracy (PCA) as compared with the state of art existing techniques.

Which aspects of soundscape can the soundscape attributes measure?

The eight perceptual attributes for soundscape assessment provided in ISO/TS 12913-2: 2018 are widely used in recent soundscape studies. Several studies across the language showed that the basic structure of the soundscape appraisal two-dimension space obtained from the attributes are robust. However, this robustness of the basic structure only means the robustness of the linguistic structure of the eight perceptual attributes, and never means those attributes cover the whole human perception of the soundscapes. Some studies suggest there are some appraisal scales which cannot be expressed in the two-dimensional appraisal space. This study discusses which aspects of soundscape can the soundscape attributes measure.

VIF-Regression Screening Ultrahigh Dimensional Feature Space

Iterative Sure Independent Screening (ISIS) was proposed for the problem of variable selection with ultrahigh dimensional feature space. Unfortunately, the ISIS method transforms the dimensionality of features from ultrahigh to ultra-low and may result in un-reliable inference when the number of important variables particularly is greater than the screening threshold. The proposed method has transformed the ultrahigh dimensionality of features to high dimension space in order to remedy of losing some information by ISIS method. The proposed method is compared with ISIS method by using real data and simulation. The results show this method is more efficient and more reliable than ISIS method.

Crop Yield Prediction using Granular SVM

Agriculture is the backbone of the Indian economy. Farming is a major source of income for many people in developing countries. Prediction of yield of crops is desirable as it can predict the income and minimise losses for the farmersunder unfavorable conditions. But predicting crop yield is a challenging task in developing countries like India. Conventionally, crop yield prediction is done using farmer’s expertise. The sustainability and productivity of a crop growing area are dependent on suitable climatic, soil, and biological conditions. So, data mining techniques based on neural networks, Neuro-Fuzzy Inference Systems, Fuzzy Logic, SMO, and Multi Linear Regression can be used for prediction. Previous work has performed yield prediction based on crop models considering only some of the environmental factors. This work uses a Support Vector Machine (SVM) to predict the crop yield under different environmental conditions that include soil, climate, and biological factors. Applying granular computing enables dividing the problem space into a sequence of subtasks. So, the hyperplane construction of SVM can be parallelized by splitting the problem space. Testing can also be parallelized. The main advantage is that linear SVM can be used to handle higher dimension space. Time complexity is reduced. Prediction using granular SVM can be parallelized using appropriate techniques like MapReduce/GPGPU. IoT-based agriculture increases crop yield by accurate prediction, automation, remote monitoring, and reducing wastage of resources. IoT-based monitoring systems can be used by farmers, researchers, and government officials to analyze crop environments and statistical information to predict crop yield. This paper proposes an IoT-based system to predict crop yield based on climatic, soil, and biological factors using parallelized granular support vector machines.