Cosine Similarity Research Articles

Analysis of Hot Topics and Evolution of Research in World-class Agricultural Universities Based on BERTopic

Abstract In this paper, the cited papers and cited patents of world-class agriculture-related universities are taken as the research objects, and the preprocessing process of the research hotspot text of agriculture-related universities is accomplished by combining the participle processing method and the TF-IDF method. To make up for the shortcomings of the semantic level of the LDA model, BERTopic is utilized to obtain the document-topic distribution probability and the topic-topic word distribution probability, and the topic hotness of different topics is calculated. Then the Glove method is used to extract the word vector features of the topic words, normalize them, calculate the word vectors by taking the distribution probability values as the weighted weights, solve the topic vectors, and then calculate the similarity of the hot topics in different time windows by the cosine similarity to explore the hot topics and the evolution of the world’s first-class agriculture-related colleges and universities research.

Research on Intelligent Development and Practice of Accounting Laboratory Courses in Colleges and Universities in the Context of Interdisciplinary Cooperation

Abstract In this paper, an innovative solution is proposed to address the problems of single content design, lack of relevance and specialized training in traditional accounting laboratory courses. In interdisciplinary cooperation, an efficient intelligent assistance model for accounting laboratory courses is constructed by combining the collaborative filtering recommendation model and K-Means clustering algorithm. The model significantly improves the recommendation accuracy and computational efficiency by optimizing the cosine similarity calculation and introducing the similarity-weighted average, combined with the enhanced dichotomous K-means algorithm. After dataset testing and practical application verification in a university, the model effectively improves the quality of personalized recommendation of teaching resources. In particular, students’ mastery of knowledge points K2 and K3 is more adequate through the model recommendation, while their mastery of K8 is significantly improved. In addition, the model successfully identifies the similarity between students, which lays the foundation for providing more personalized learning resources. Overall, the model solves the traditional problem of accounting laboratory courses and provides a practical methodology for intelligent recommendation of teaching resources in universities.

Implementation of Cosine Similarity Algorithm on Omnibus Law Drafting

Implementation of Cosine Similarity Algorithm on Omnibus Law Drafting

DARK: Few-Shot Remote-Sensing Colorization Using Label-Conditioned Color Injection

Satellite image colorization is a broad challenging problem in the domain of remote sensing (RS) having huge potential applications. The problem becomes even more complicated under the few-shot setting yet it has barely been studied to date. In this paper, we propose a colorization framework for the RS scene for synthesizing optical images from their panchromatic (PAN) counterparts using color injection and attention fusion mechanism. Our proposed model ensures that the synthesized optical images are coherent with the structural variability of panchromatic images while constraining the realistic appearance in the optical domain from a few training image pairs. To accomplish the same, we introduce a novel Dual Attention fusion of Receptive Kernels (DARK) which considers the spatial nuances along with color injection conditioned on prior label allocation. DARK is a multi spectral-spatial feature generator that selectively accentuates important cross-spatial features based on attention fusion. We also employ a prior distribution constraint on color embedding generation for introducing vibrant yet diverse variance in a color generation. Our approach achieves state-of-the-art results on the publicly available EuroSAT and PatternNet datasets while demonstrating significant speedups. We showcase our results quantitatively by comparing the PSNR, mean squared error(MSE), and cosine similarity of generated images and qualitatively via visual perception.

3-D Sharpened Cosine Similarity Operation for Hyperspectral Image Classification

3-D Sharpened Cosine Similarity Operation for Hyperspectral Image Classification

Optimizing Quality Tool Selection with Cosine Similarity for Continuous Improvement

Optimizing Quality Tool Selection with Cosine Similarity for Continuous Improvement

Selection of Distance Measure for Visual and Long Wave Infrared Image Region Similarity using CNN Features

Similarity computation between images or image regions is a necessary precursor for several vision-based applications, such as retrieval, registration, change detection etc. A two-channel convolutional neural network architecture is designed to retrieve an appropriate visual (VS) image representative from the repository, given as query a long-wave infrared (LWIR) image patch of the same region. Both the VS and LWIR image regions are described using pretrained convolution neural network models and images are ranked by computing the dis/similarity between the feature vectors. It is essential to evaluate and identify the suitable combination of feature descriptor and distance measure, when applied to LWIR and visual image similarity, as pre-trained CNNs such as VGG16, VGG19, MobileNet are not trained for LWIR images. RoadScene dataset which contains a pair of aligned long-wave infrared and visible images is used and performance of CNN features and distance measures is objectively evaluated using computation time, number of patches in top 5 and also in computing how close the retrieved LWIR patch is to the visual patch. Results demonstrate that Cosine similarity measure is relatively better when compared to all the other distance measures in addressing the spectral variations.

DyCARS: A dynamic context-aware recommendation system.

Dynamic recommendation systems aim to achieve real-time updates and dynamic migration of user interests, primarily utilizing user-item interaction sequences with timestamps to capture the dynamic changes in user interests and item attributes. Recent research has mainly centered on two aspects. First, it involves modeling the dynamic interaction relationships between users and items using dynamic graphs. Second, it focuses on mining their long-term and short-term interaction patterns. This is achieved through the joint learning of static and dynamic embeddings for both users and items. Although most existing methods have achieved some success in modeling the historical interaction sequences between users and items, there is still room for improvement, particularly in terms of modeling the long-term dependency structures of dynamic interaction histories and extracting the most relevant delayed interaction patterns. To address this issue, we proposed a Dynamic Context-Aware Recommendation System for dynamic recommendation. Specifically, our model is built on a dynamic graph and utilizes the static embeddings of recent user-item interactions as dynamic context. Additionally, we constructed a Gated Multi-Layer Perceptron encoder to capture the long-term dependency structure in the dynamic interaction history and extract high-level features. Then, we introduced an Attention Pooling network to learn similarity scores between high-level features in the user-item dynamic interaction history. By calculating bidirectional attention weights, we extracted the most relevant delayed interaction patterns from the historical sequence to predict the dynamic embeddings of users and items. Additionally, we proposed a loss function called the Pairwise Cosine Similarity loss for dynamic recommendation to jointly optimize the static and dynamic embeddings of two types of nodes. Finally, extensive experiments on two real-world datasets, LastFM, and the Global Terrorism Database showed that our model achieves consistent improvements over state-of-the-art baselines.

Privacy-Preserving Federated Learning Against Label-Flipping Attacks on Non-IID Data

Federated learning (FL) has attracted widespread attention in the Internet of Things domain recently. With FL, multiple distributed devices can cooperatively train a global model by transmitting model updates without disclosing the original data. However, the distributed nature of FL makes it vulnerable to data poisoning attacks. In practice, malicious clients can launch the label-flipping attack by simply tampering with the labels of local data, thus causing the global model to misclassify the samples of a selected class as the target class. Although some defense mechanisms have been proposed, they rely on specific assumptions about data distribution, and their performance degrades significantly when the data on clients are non-IID. Besides, most existing methods require clients to upload model updates in plaintext so that the server can identify and remove the malicious updates. But, direct transmission of model updates may still reveal private information. Considering these issues, we develop a label-flipping-robust and privacy-preserving FL (LFR-PPFL) algorithm, which is applicable to both IID and non-IID data. We first propose a detection method based on temporal analysis on cosine similarity to distinguish malicious clients from benign clients. Then, we propose a privacy-preserving computation protocol based on homomorphic encryption to implement this detection method and perform federated aggregation while protecting the privacy of clients. Besides, a detailed theoretical analysis is given to demonstrate the privacy guarantee of the proposed protocol. Experimental results on real-world datasets show that the proposed algorithm can effectively defend against label-flipping attacks under various data distributions.

Research on the Problems of Teaching Traditional Music Basic Theory in the Information Age

Abstract In the era of informationization, the inheritance and development of traditional music are facing challenges and opportunities. This study analyzes the problems faced by teaching the basic theory of traditional music and proposes a coping strategy for constructing an interactive teaching model and platform for informatization. In the interactive teaching model, P2P technology is used to design and construct the interactive behavior model and interest analysis module, and fuzzy theory is used to precisely set the key information weights. The informationized teaching platform is designed with AI technology, adopting the recommendation strategy of a hybrid recommendation model combined with collaborative filtering, using cosine similarity to calculate the content similarity, and combining with artificial intelligence technology to realize music teaching quality assessment. In the traditional music basic theory teaching practice, S High School is used as a practice site, and experimental and control classes are set up to conduct teaching. The experimental class perceived negative behaviors such as “harshness” and “dissatisfaction” as lower than the control class, and the pass rate and excellence rate of the post-test of traditional music basic theory reached 92.5% and 36.9%, respectively. The mean values of all other music core literacy dimensions except for knowledge of opera and general knowledge of musical instruments were higher than those of the control class, showing significant differences (P<0.05).

Research on Efficient English Teaching Model Combining Corpus Microtext Analysis

Abstract Aiming at the goal of improving the quality of English teaching and combining the characteristics of English teaching, this paper constructs an efficient English discursive literacy teaching based on corpus micro-text analysis. Utilizing the unique advantages of online corpus to assist English teaching, it proposes the targeted teaching of English metaphorical vocabulary and combines the Word2Vec word embedding model and the cosine similarity algorithm to propose a model for analyzing the quality of English text coherence. Based on the characteristics of writing vocabulary in microtexts of the corpus, the analysis integrates the dimensional features of English writing vocabulary teaching. The effectiveness of the corpus-based efficient teaching model for English literacy teaching is verified through teaching practice, specifically utilizing the average values of pre- and post-tested vocabulary, sentences, and chapters of the two classes as comparison data. The results proved that the average value of the pre-test scores of all students in the experimental class is 7.78, and the average value of the post-test is 10.72, which is 2.94 higher than the pre-test. The English teaching model in colleges and universities based on corpus micro-text analysis can significantly improve the English writing level of students and achieve efficient English teaching.

Innovation of Teaching Methods of Civic and Political Education in Colleges and Universities with the Assistance of Artificial Intelligence

Abstract Artificial intelligence empowers education and promotes its integration and innovation, enabling the use of learning analytics to accurately diagnose the learning situation and implement precise teaching. In this paper, we have leveraged artificial intelligence to innovate traditional teaching methods, developed a wisdom teaching model specifically for college civic education, and concentrated on the precise application of wisdom teaching in this context. First, it creates student profiles based on academic data and then provides personalized recommendations for learning resources by combining a collaborative filtering algorithm and an improved K-means algorithm. Finally, we analyze the actual impact of the teaching intervention by comparing the student portraits before and after the intervention. When the number of neighbors K takes the same value, the improved cosine similarity calculation method outperforms other calculation methods, and the RMSE, MAE, accuracy, recall, and F1 values are 0.14, 0.24, 0.99, 0.74, and 0.85 when K = 60, respectively, which indicates that the algorithm improvement in this paper is practical. This paper categorizes the sample students into four groups based on their learning behavior and outcomes: potential, excellent, marginal, and hardworking. All these student types have demonstrated some progress following the teaching intervention, confirming the efficacy of this paper’s artificial intelligence-based teaching approach for college civic education. Simultaneously, the created profiles of these student types establish the groundwork for future personalized and accurate instruction.

Biometric lock with facial recognition implemented with deep learning techniques

The increased criminal activity associated with unauthorized entry into facilities has become a global concern. Traditional mechanical locks suffer from drawbacks such as key loss, theft, duplication risks, and time-consuming operation. Therefore, biometrics has been explored as a key to accessing a restricted area. However, some challenges still need to be solved in developing such systems, including user registration, response speed, maintainability, and the ability to distinguish between real and fake individuals. This paper proposes and develops a biometric lock system (BLS) whose opening is performed by recognizing a person?s face. It solves the challenges of re-training, antispoofing, real-time response, and works inside an embedding system. The proposed BLS overcomes these challenges using a pre-trained network called FaceNet for feature extraction and coding into 128-dimensional vectors.We use the characteristic vector provided by FaceNet and a cosine distance to recognize the persons. It also incorporates ResNet18 + remote photoplethysmography (rPPG) to avoid spoofing. The architecture was implemented in a BLS, demonstrating an impressive false acceptance rate of 0% under varying lighting conditions, with an average response time of 1.68 seconds from facial detection to door opening. The BLS has easy maintainable devices, providing enhanced security by accurately identifying individuals and preventing unauthorized access. The system can distinguish between real and fake people without using specialized hardware. Making it a versatile solution suitable for homes, offices, and commercial spaces. The results underscore the potential efficacy of our proposed BLS in mitigating security challenges related to unwarranted access to restricted facilities.

A Healthcare Recommender System Framework

After the pandemic hit every part of the world, healthcare awareness is slowly rising among every human being, especially leaders of each country. Due to a shortage of manpower in the healthcare industry, patients tend to search the internet for some self diagnoses. This way is extremely dangerous as patients might end up using the wrong treatment such as taking the wrong medication to treat their sickness since there are so many different remedies posted on the internet without valid recognition from the healthcare professionals. To aid in overcoming this problem, this research will be building a Healthcare Recommender System. The goal of a Healthcare Recommender System (HRS) aims to supply its user with medical information that is meant to be highly relevant and tailored to an individuals need. Hence, this paper gives an overview of various recommender systems, datasets employed, and evaluation metrics used in the healthcare system. In addition, we propose the framework for the HRS to capture user input on their condition and recommend the next course of action. The steps involved in our recommender system includes choosing the dataset and techniques, data cleaning and preprocessing, building the recommender system, training the recommender engine, and finally performing the prediction. We generate the accuracy of prediction and analyze with some results. From the experimental results, Cosine Similarity has the highest accuracy compared to Jaccard Similarity and Euclidean Distance.

A Healthcare Recommender System Framework

After the pandemic hit every part of the world, healthcare awareness is slowly rising among every human being, especially leaders of each country. Due to a shortage of manpower in the healthcare industry, patients tend to search the internet for some self diagnoses. This way is extremely dangerous as patients might end up using the wrong treatment such as taking the wrong medication to treat their sickness since there are so many different remedies posted on the internet without valid recognition from the healthcare professionals. To aid in overcoming this problem, this research will be building a Healthcare Recommender System. The goal of a Healthcare Recommender System (HRS) aims to supply its user with medical information that is meant to be highly relevant and tailored to an individuals need. Hence, this paper gives an overview of various recommender systems, datasets employed, and evaluation metrics used in the healthcare system. In addition, we propose the framework for the HRS to capture user input on their condition and recommend the next course of action. The steps involved in our recommender system includes choosing the dataset and techniques, data cleaning and preprocessing, building the recommender system, training the recommender engine, and finally performing the prediction. We generate the accuracy of prediction and analyze with some results. From the experimental results, Cosine Similarity has the highest accuracy compared to Jaccard Similarity and Euclidean Distance.

Revisiting Probabilistic Relation Analysis: Using Probabilistic Relation Graphs for Relational Similarity Analysis of Words in Short Texts

Relation graphs provide useful tools for structural and relational analyses of highly complex multi-component systems. Probabilistic relation graph models can represent relations between system components by their probabilistic links. These graph types have been widely used for the graphical representation of Markov models and bigram probabilities. This study presents an implication of relational similarities within probabilistic graph models of textual entries. The article discusses several utilization examples of two fundamental similarity measures in the probabilistic analysis of short texts. To this end, the construction of probabilistic graph models by using bigram probability matrices of textual entries is illustrated, and vector spaces of input word-vectors and output word-vectors are formed. In this vector space, the utilization of cosine similarity and mean squared error measures are demonstrated to evaluate the probabilistic relational similarity between lexeme pairs in short texts. By using probabilistic relation graphs of the short texts, relational interchangeability analyses of lexeme pairs are conducted, and confidence index parameters are defined to express the reliability of these analyses. Potential applications of these graphs in language processing and linguistics are discussed on the basis of the analysis results of example texts. The performance of the applied similarity measures is evaluated in comparison to the similarity index of the word2vec language model. Results of the comparative study in one of the illustrative examples reveal that synonyms with 0.18157 word2vec similarity value scored 1.0 cosine similarity value according to the proposed method.

Skin Lesion Classification Using Convolutional Neural Network and ABCD Rule

Skin cancer, which can occur in any part of the human skin, is one of the common and serious types of cancer. Accurate diagnosis and segmentation of lesions are crutial to the early diagnosis. Computer-aided diagnosis make important contributions to help doctors in the diagnosis of cancer from skin images. The most important factor for such systems to reveal the accurate results is the correct feature extraction. In this study, a model for the classification of seven types of skin lesions was developed by combining the features of CNN-based feature extraction and the ABCD rule, which is widely used in the clinic. The model was evaluated on HAM10000 well-known dataset. The classification results obtained with different combinations of features and machine learning algorithms were compared. According to the results, the best classification accuracy was obtained with the Cosine Similarity Classifier with 96.4% when the features determined by CNN and the features in the ABCD rule were used together.

Research on E-Commerce Commodity Demand Forecasting and Inventory Optimization Based on Time Series Forecasting Model

In this study, 1996 sets of time series data are extracted from the three dimensions of merchant, warehouse and commodity, and the outliers are removed using the 3σ criterion. Through visual analysis, it was found that home improvement building materials and food and beverages were popular commodities. Then ARIMA and LSTM models were used to forecast the demand. Through 1-wmape index evaluation, LSTM model was found to perform better. Finally, the K-Means cluster analysis is carried out based on the categories of goods, merchants and warehouses, and the cluster number is selected as 70. The cosine similarity method is used to match the newly added dimensions to the category of the cluster in problem 1, and the ARIMA model is used for prediction. The results show that the cosine similarity method can accurately predict the commodity demand, the 1-wmape index is 0.801, and the residual plot presents randomness. The data of November 2022 is added to the original data, and the missing data is processed by linear interpolation. The LSTM model was used to forecast the increase of the total demand during Double 11. By setting a periodic time series with a step size of 30, the model achieves a good accuracy in predicting the demand in June, with a 1-wmape index of 0.829. To sum up, this study comprehensively considers three dimensions of merchants, warehouses and commodities, and uses different models to make accurate predictions. The ARIMA and LSTM models are adopted, while the promotion characteristics are considered and the appropriate LSTM parameters are selected. The research results provide practical forecasting tools and methods for supply chain management.

기술변화가 고용과 임금에 미치는 영향

This study links and analyzes textual data on patents and occupations to create a patent volume by occupation and analyzes the effect of skill characteristics proxied by patent volume on employment and wages. We used CPC code descriptions, and Occupational Dictionary descriptions, linking US SOC, IOC, KSOC. Using FastText as a word embedding method, we calculated the cosine similarity of O*NET and CPC texts to create a patent volume variable by occupation. The analysis shows that the effect of technology on employment and wage changes is statistically significant (-) in the panel fixed effects model. While the employment effect is largely non-linear, i.e., decreasing in the short run and increasing in the medium and long run, the wage effect shows a negative linear relationship. These findings are largely in line with existing international studies in this area and confirm the general finding that the effects of technological change are not uniform or unidimensional and require policy responses that take into account different contexts and dimensions.

CalAttnMix: An instance-attention based data augmentation method for calcification detection in mammograms

Data augmentation is widely used as a pre-processing step to solve limited training data and class imbalance problems in deep-learning based calcification detection methods. However, current augmentation methods in breast mammograms fail to guarantee the class balance and diversity of generated calcifications, especially that randomly fusing instances with specified regions results in unrealistic images, further degrading the performances. Here, we propose an instance-attention based data augmentation method, named Calcification-Attention-Mix (CalAttnMix), which adaptively transfers multiple source calcification instances into diverse breast tissue backgrounds, aiding breast calcification detection. First, to seek feasible breast tissue candidate backgrounds, an Augmentation Region Proposal Mechanism (ARPM) is devised based on morphology constraints of calcifications. Then, without segmentation masks of calcification foregrounds, we design a transformer-based Foreground Prediction Module (FPM), along with a multi-task learning strategy, to localize instances accurately in a weakly-supervised manner. Finally, to generate diverse instances, an attention mix module is introduced to dynamically mix instances with background candidates, simulating various breast tissue surroundings. By focusing on the minor class, CalAttnMix generates diverse and balanced datasets for calcification detection. Experiments on the public and clinical dataset show that CalAttnMix achieves the lowest FID 22.152 and cosine similarity score 0.304, respectively, confirming the realism and diversity of generated images. CalAttnMix outperforms SOTA augmentation methods by 3.40% / 2.30% and 1.40% / 5.00% in the average recall/mAP, demonstrating the best detection results.