Volume 2

Traditional edge detection operators are usually applied on edge detection in 2D image processing. However, the edge detection system equipped with simple operators has many disadvantages, such as high sensitivity to noise and neglect of significant edge details. This work proposed a method to enhance edge detection with convolutional neural net-work. To overcome the shortcomings of the system using simple edge detection operators in 2D image processing, an edge detection system using convolutional neural network was developed with Python language. In the convolutional neural network, two convolutional layers were designed to extract 2D image features that were relative to edge information. Then a normalization layer was applied to normalize the convoluted output. After that, pre-processing was utilized to denoise and smooth the image input. The final step was edge detection using traditional operators. Experiments were also implemented to verify the improvement of the plugin of the three-layer convolutional neural network in the de-signed edge detection system. Relative frequencies were utilized to quantify the edge de-tection performance. Results showed that the involvement of convolutional neural net-work could strengthen edge detection operators’ performance obviously in computer vi-sion.

Show more

The rising demand for beauty and healthy lifestyles makes the global beauty and spas salons market promising. How to choose appropriate salons for a specific customer becomes an issue and is underexplored. In this paper, we built two types of personalized recommendation system models, the pure model-based collaborative filtering model and the LightFM model (a kind of hybrid recommendation system model) to make recommendations for beauty and spas salons based on Yelp Dataset. The results showed the LightFM model had a better performance. Besides, by applying aspect-based sentimental analysis to extract new features from customer reviews, we further improved the performance of the LightFM model.

Show more

With the outbreak of COVID-19, wearing masks has become a hot topic again. In public places, Not wearing a mask correctly has almost the same harm as not wearing a mask. Mask detection is an extension of face detection. Therefore, it is of great practical signifi-cance to design a system that can correctly identify whether pedestrians wear masks cor-rectly. The face recognition technology based on neural network has been relatively ma-ture, but there is still a lack of work related to mask recognition, especially whether to wear masks correctly. This is not only the promotion of a two-classification problem to a three-classification problem, but also faces many practical problems, including data set acquisition, in this paper, mask recognition is divided into a multi-stage work. In the first link, Yolo network is used to recognize facial areas, and in the second link, RESNET is used to realize mask recognition. Finally, a layer of RESNET network is improved to achieve higher recognition accuracy.

Show more

A neurodevelopmental disorder named autism spectrum disorder (ASD) is challenging to diagnose. The prevailing diagnostic manner is based merely on the behavioral measure-ment with a high tendency of misdiagnosis. People require an advanced method to make more quantitative diagnosis. In this paper, two deep learning architectures were explored with the machine learning methods. The Mixup method was used to augment the original functional Magnetic Resonance Imaging data. Features of the data extracted by two dif-ferent kinds of autoencoders which are Sparse Autoencoder and Variational Autoencoder were used as inputs of two deep neural networks functioning as classifiers respectively. The models can classify patients with ASD from typical control subjects with the accura-cy of 75.5% and 75.2% respectively, which outperformed the other state-of-the-art meth-od by 4.7% and 4.4%. The further significance of this project is to help develop our per-ception of the neurobiological foundation of the ASD.

Show more

We propose a multi-model pipeline for image-caption matching tasks on Wikipedia-based dataset which leverages object-detection technique and attention mechanism to achieve fine-grained matching between textual representation and image representation. Different from the prior research, we not only evaluate our pipeline effectiveness on common benchmark dataset such as MS-COCO and Flickr30k, but also a new dataset that is de-rived from Wikipedia which is rich in natural entities and abstract concepts. Our findings show: 1) our model pipeline improves R@1 by 113.4%, R@3 by 86.1%, and R@5 by 114.4% compared to the original pipeline provided by the Wikipedia-based dataset. 2) our model pipeline has close to the state-of-the-art performance in common benchmark dataset including Flickr30k and MS-COCO. 3) images that are from Wikipedia creates bigger challenges for models to understand compares to MS-COCO or Flickr30k due to the abstract concepts and broad topics covered by Wikipedia.

Show more

Facing the increasing need of data transmission security, encrypting data through cryptography is a key solution. In an industrial Internet of things, cryptography based on quantum key distribution provides perfect secrecy with lower resource requirements of both computational power and storage compared with traditional cryptography. To explore how to deploy this technology, this paper proposes an industrial Internet of things network architecture embedded with quantum key distribution systems, combines it with the noise model of spontaneous Raman scattering and the evaluation model of quantum key distribution systems theoretically, simulates the performance in a normally used industrial environment, and works out instructions to the deployment of the raised architecture. The results also show that a better choice to avoid performance descending is to duplicate classical channels and quantum channels with the same direction instead of moving classical channels backward, while noises have the strongest influence at the transmission distance of 25 km.

Show more

In this paperwork, the Apriori algorithm can be used to describe the association between data in a database. This paper will use the demographic and financial information of the users in the data set to analyze the user profile as the basic information. This work will fit the model by optimizing the Apriori algorithm, generating a combination of two or more associated variables, and then using the decision tree to learn and output the final model. The main contribution of this paper is to explore the optimization of Apriori and the ef-fect of Apriori's combined model with the decision tree in predicting user buying behavior. Therefore, the conclusion can be drawn that in the sale of deposit products, if the custom-er has the above characteristics, such as no credit fault, no loan, housing loan, with a col-lege degree, marital status is single, and act as an administrator. The people mentioned above may be more interested in the product. In other words, it is more likely to sell suc-cessfully. In future work, in addition to deposit products, more customer buying infor-mation is highly needed to extend predictable results to broader anger.

Show more

This research analyzed the relationship between multiple elements of a book's classifica-tion through natural language processing and machine learning. This paper used SVM and KNN to classify books according to the titles and author respectively. Also, books are categorized by summary through Decision Tree, Naïve Bayes and BERT. In the end, this work compared effects of these methods. Our findings show that 1) books have different levels of categorical characteristics in various parts of the book 2) through combining the title, author, and summary of the book, more accurate classification results were obtained 3) BERT achieved more accurate recognition compared to a variety of other algorithms used.

Show more

With the advent of the Internet era, online shopping has become an integral part of people’s life. In order to perform precision marketing, more and more e-commerce platforms are trying to predict users’ repurchase behaviors by collecting massive user behavior data. Although the traditional single-model prediction method is mature, it is still difficult to improve the accuracy of prediction. Based on the real user behavior data of Tmall, this paper focuses on comparing and exploring the help of different algorithm fusion methods to improve the model prediction effect. The under-sampling method is introduced for sample equalization processing. User behavior features are constructed from three aspects which are user, merchant and user-merchant interaction. Taking AUC value as evaluation method, Soft-Voting and Stacking model fusion methods are used to integrate logistics regression, KNN, XGBoost and RandomForest. And the prediction results is produced based on stratified 5-fold cross-validation. The experimental results show that the fusion model can effectively improve the prediction effect, and the AUC value is raised by 0.2%~4% compared with the single model. The AUC value of Soft-Voting increases by approximately 0.4% after it is weighted.

Show more

This article provided an introduction of applying reinforcement learning to games, including board games and video games like Backgammon, Go, and Dota2. The reason for choosing reinforcement learning to solve game problems was analyzed. The article also reviewed the reinforcement learning technique and introduced two optimizing learning methods, Temporal Difference learning and Q learning. Then, three important cases of using reinforcement learning to reach high level game skill, TD-gammon, AlphaGo, and OpenAI Five, were introduced. In the end, the future possibility of applying reinforcement learning in broader way was analyzed.

Show more