In present days, Multi Label Text Classification (MLTC) has become an important research area in Natural Language Processing (NLP). Existing models of MLTC have been facing domain specificity and fine-tuning challenge...
详细信息
ISBN:
(数字)9798350370249
ISBN:
(纸本)9798350370270
In present days, Multi Label Text Classification (MLTC) has become an important research area in Natural Language Processing (NLP). Existing models of MLTC have been facing domain specificity and fine-tuning challenges. Hence a model is developed to overcome the above said challenges by introducing Prompt engineering, GPT FineTuning methods and Pipeline algorithms. The experimental results are obtained based on the present Generative pre trained data produces labels for various text documents and files with an exceptional accuracy of 85%. The outcomes of the experiments reveal notable enhancements over the baseline models.
In recent years, self-training has become more and more popular in model learning because it only requires a small amount of labeled data to reach high classification accuracy. Since the conventional self-training doe...
In recent years, self-training has become more and more popular in model learning because it only requires a small amount of labeled data to reach high classification accuracy. Since the conventional self-training does not perform in a class-specific manner, class imbalance problem happens frequently as self-training tends to select more pseudo labeled samples from easy classes. It becomes even worse when training dataset is originally class-imbalanced. In this paper, classwise self-paced self-training is proposed to solve this problem. A classwise pseudo-labeling method ensures that sufficient data are selected for each class in a self-pace manner. To avoid increasing number of noisy samples, two indicators, confidence score and uncertainty score, are proposed based on dropout prediction for noisy label filtering. Experimental results demonstrate that the proposed method achieves very competitive performance on CIFAR-10 and ISIC2018 datasets. Keyword: Deep Learning, Image Classification, Semi-Supervised Learning, Self-Training
This paper proposes to use both audio input and subject information to predict the personalized preference of two audio segments with the same content in different qualities. A siamese network is used to compare the i...
详细信息
This paper presents a security system that uses the built-in Virtual Private Networks (VPN) mechanism on Android to monitor the Uniform Resource Locators (URLs) linked to when apps are executed. Then, the proposed sys...
This paper presents a security system that uses the built-in Virtual Private Networks (VPN) mechanism on Android to monitor the Uniform Resource Locators (URLs) linked to when apps are executed. Then, the proposed system sends the URLs to VirusTotal, a third-party security website for evaluation. Finally, the security score of each URL is presented to the smartphone user, providing them the choice to decide whether to remove an app with security concerns.
We propose an end-to-end learned image data hiding framework that embeds and extracts secrets in the latent representations of a generic neural compressor. By leveraging a perceptual loss function in conjunction with ...
详细信息
This paper presents a scoring system for evaluating makeup paper pictures using computer vision feature recognition. In traditional makeup scoring systems, the assessment is conducted by human reviewers, leading to po...
详细信息
A substantial and rising number of patients suffer from cardiovascular diseases, including heart attacks, heart failure, and other related illnesses. This case surge places increasing pressure on healthcare profession...
A substantial and rising number of patients suffer from cardiovascular diseases, including heart attacks, heart failure, and other related illnesses. This case surge places increasing pressure on healthcare professionals and administrators while patients grapple with growing medical costs. To address these challenges, an automated system is necessary within the healthcare sector. In this paper, we present a cardiovascular health monitoring system that incorporates Machine Learning techniques. The study employed feature selection techniques on a secondary dataset. The study has used a nature-inspired technique, namely the Clonal Selection Algorithm (CSA) and Maximum Relevance Minimum Redundancy (mRMR) technique, to identify the most prominent feature in the detection of cardio-vascular illness. This approach was combined with a collection of Machine Learning classifiers. A group of experimental data has been listed to assess the suggested model's effectiveness. The research findings indicated a maximum accuracy rate of 100% when employing the proposed algorithms and orientations. The study also discusses the performance analysis for CSA and mRMR using a set of performance evaluation matrices. Based on the obtained results, it can be inferred that the suggested model will likely exhibit a high level of effectiveness in identifying cardiovascular diseases.
This paper proposes to use both audio input and subject information to predict the personalized preference of two audio segments with the same content in different qualities. A siamese network is used to compare two i...
This paper proposes to use both audio input and subject information to predict the personalized preference of two audio segments with the same content in different qualities. A siamese network is used to compare two inputs and predict the preference. Several different structures for each side of the siamese network are investigated. The baseline structure which uses only audio information involves using a pretrained audio encoder followed by fully connected layers. In several different proposed structures, the approach of concatenating subject information with audio embedding before feeding it into fully connected layers outperforms the baseline model the most, resulting in an increase in overall accuracy from 77.56% to 78.04%. Experimental results also demonstrate that utilizing the complete set of subject information, which includes age, gender, and headphone/earphone specifications such as impedance, frequency response range, and sensitivity, is more effective than using a subset of this information.
This study introduces "CapSight,"a system for enhancing social interaction in visually impaired individuals. It includes a single-board computer, camera, vibration motors, and Bluetooth bone conduction headp...
详细信息
Deep learning-based sign-language recognition usually requires abundant training videos. This research considers generating valid sign-language data for training the recognition models. MediaPipe is used to acquire th...
Deep learning-based sign-language recognition usually requires abundant training videos. This research considers generating valid sign-language data for training the recognition models. MediaPipe is used to acquire the hand skeleton from the sign-language video. Then we analyze several hand skeleton adjustment policies with color-weighting strategies and generate hand masks to simulate different hands. Since miss detections of hands may happen due to motion blurring caused by rapid hand movements, we incorporate optical flows to ensure that the hand movement information is retained in each frame. We employ different spatial and temporal data augmentation strategies to simulate varying hand sizes, filming angles, and hand speeds. The experimental results show that the proposed method improves the accuracy of sign-language recognition in the American Sign Language dataset.
暂无评论