A Deep Learning Approach to Nightfire Detection Based on Low-light Satellite
Yue Wang, Ye Ni, Xutao Li and Yunming Ye, Department of Computer Science, Harbin Institute of Technology, Shenzhen, China
Wildfires are a serious disaster, which often cause severe damages to forests and plants. Without an early detection and suitable control action, a small wildfire could grow into a big and serious one. The problem is especially fatal at night, as firefighters in general miss the chance to detect the wildfires in the very first few hours. Low-light satellites, which take pictures at night, offer an opportunity to detect night fire timely. However, previous studies identify night fires based on threshold methods or conventional machine learning approaches, which are not robust and accurate enough. In this paper, we develop a new deep learning approach, which determines night fire locations by a pixel-level classification on low-light remote sensing image. Experimental results on VIIRS data demonstrate the superiority and effectiveness of the proposed method, which outperforms conventional threshold and machine learning approaches.
night fire detection, pixel segmentation, low-light satellite image
Context-aware Short-term Interest First Model for Session-based Recommendation
Haomei Duan and Jinghua Zhu, School of Computer Science and Technology, Heilongjiang University, Harbin, China
In the case that user profiles are not available, the recommendation based on anonymous session is particularly important, which main aim is to predict the items that the user may click at the next moment based on the users access sequence over a while. In recent years, with the development of recurrent neural network, attention mechanism, and graph neural network, the performance of session-based recommendation has been greatly improved. However, the previous methods did not comprehensively consider the context dependencies and short-term interests first of the session. Therefore, we propose a context-aware short-term interest first model (CASIF). In CASIF, we dynamically construct a graph structure for session sequences and capture rich context dependencies via graph neural network (GNN), latent feature vectors are captured as inputs of the next step. Then we build the short-term interest first module, which can to capture the users general interests from the session in the context of long-term memory, at the same time get the users current interests from the item of the last click. In the end, the short-term and long-term interests are combined as the final interest and multiplied by the candidate vector to obtain the recommendation probability.
recommendation, session-based, context-aware, neural, network, attention
User Characteristics of Olympic Gold Medallists on Instagram: A Quantitative Analysis of Rio2016
Amirhosein Bodaghi, Federal University of Rio de Janeiro, Department of Computing Science, Centre of Mathematical and Natural Sciences – CCMN, Rio de Janeiro, Brazil
The purpose of this study is to examine Olympic champions’ characteristics on Instagram to first understand whether differences exist between male and female athletes and then to find possible correlations between these characteristics. We utilized a content analytic method to analyse Olympic gold medallists’ photographs on Instagram. By this way we fetched data from Instagram pages of all those Rio2016 Olympic gold medallists who had their account publically available. The analysis of data revealed the existence of a positive monotonic relationship between the ratio of following/follower and the ratio of engagement/follower for men gold medallists, and a strong negative monotonic relationship between age and ratio of self-presenting post of both men and women gold medallists which even take a linear form for men. These findings aligned with the relative theories and literature may come together to help the athletes to manage and expand their personal brand in social media.
Instagram, self-presenting, user characteristics, Olympics, gold medallists
Towards Adversarial Genetic Text Generation
Deniz Kavi, The KoÃ§ School, Turkey
Text generation is the task of generating natural language, and producing outputs similar to or better than human texts. Due to deep learning’s recent success in the field of natural language processing, computer generated text has come closer to becoming indistinguishable to human writing. Genetic Algorithms have not been as popular in the field of text generation. We propose a genetic algorithm combined with text classification and clustering models which automatically grade the texts generated by the genetic algorithm. The genetic algorithm is given poorly generated texts from a Markov chain, these texts are then graded by a text classifier and a text clustering model. We then apply crossover to pairs of texts, with emphasis on those that received higher grades. Changes to the grading system and further improvements to the genetic algorithm are to be the focus of future research.
Valences Estimation for Spanish Sentiment Analysis using A Genetic Algorithm
Kevin Mejía1 and Yulia Ledeneva2 and René García3, 1Autonomus University of the State of Mexico, 2Ph. D. Autonomous University of the State of Mexico, 3Ph. D. Autonomous University of the State of Mexico
The analysis of opinions, in microblogs such as Twitter, has been a task that has acquired great interest due the large number of unstructured opinions that are not analyzed automatically. To address the above Sentiment Analysis (SA) is applied. In SA are three approaches: through lexicons, through machine learning, and a combination of the above called hybrid approach. This article presents a method which automatically estimates the valences of the words of a Spanish language lexicon using a Genetic Algorithm (GA), using such valences as training characteristics for Support Vector Machines (SVM). The proposed method was tested in the corpus of opinions in Spanish (COST). Evaluation has been carried out on three main measures: precision, recall and harmonic measures between the previous ones (FMeasure). The results obtained from the experiments carried out with the implemented method showed a great improvement in the classification task.
Genetic Algorithm, lexicons, machine learning, support vector machines, hybrid approach
Supervised Machine Learning Approaches for Sentiment Analysis on a Movie Review
Ojonukpe S. Egwuche1, Micheal O. Ajinaja2, Kolawole O. Adekunle3 & Israel D. Haruna4, 1Department of Computer Science, Federal Polytechnic, Ile-Oluji, Ondo State, Nigeria, 2Department of Computer Science, Federal Polytechnic, Ile-Oluji, Ondo State, Nigeria, 3Department of Computer Science, Federal Polytecnic, Ile-Oluji, Ondo State, 4Nigeria Road and Building Research Institute, Ota, Ogun State, Nigeria
Feedback from consumers/customers in forms of reviews provides a pool of ideas that are of immense importance to the promotional policies of any business.Developments in Information Technology have made personal blogs and online review sites possible as opinion resources for customers to access and assess the opinions of others to decide whether to buy a product or not.E-commerce websites such as Amazon, Alibaba, Jumia and social media websites such as Twitter, Facebook, etc. are widely used for effectivecommunication of viewpoints. Assignment of sentiments either positive or negative can assistusers to make informed decisions in their product selection and companies to understand their customers. Sentiment analysis is a complex problem which can be solved with either machine learning techniques with labeled data and unsupervised machine learning technique with unlabeled data.
Sentiment analysis, Movie Review Mining, Machine Learning.
The Impact of CrowdSourcing on the Performance of Concatination of Word2Vec and Glove Algorithms
Mohammad Jafarabad, Department of Computer Engineering, Qom University, Iran
Deep learning algorithms have been effective in recognizing lexical similarity. In cases where there is not enough standard labeled primary data, the volume of data can be increased by crowdsourcing. For machine learning tasks, they also combine the results of crowdsourcing, so that we have more and more accurate gold data. In this study, we combined crowdsourcing with deep learning. Crowdsourcing did the labeling process for us, and we achieved good accuracy in identifying pairs of entities.
crowdsourcing, Deep learning, gold data, labeling, word2vec.
A color image blind digital watermarking algorithm based on QR code
Xuecheng Gong and Wanggen Gong, School of Computer and Information, Anhui Normal University, Anhui Wuhu, China
The current color image digital watermarking algorithm has the problem of low robustness. Aiming at this problem, a color image blind digital watermarking algorithm based on QR code is proposed. The algorithm combines Discrete Wavelet Transform (DWT) and Discrete Cosine Transform (DCT). First, the color image was converted from RGB space to YCbCr space, and the Y component was extracted and the secondlevel discrete wavelet transform is performed; secondly, the LL2 subband was divided into blocks and carried out discrete cosine transform; finally, used the embedding method to embed the Arnold transform watermark information into the block. The experimental results show that the PSNR of the color image embedded with the QR code is 56.7159 without being attacked. After being attacked, its PSNR is more than 30dB and NC is more than 0.95. It is proved that the algorithm has good robustness and can achieve blind watermark extraction.
QR Code, Color Image, Arnold Transform, DWT.