this paper presents a novel method for SAR deception jamming recognition based on feature map learning. According to different characteristics of the true echo and deceptive jamming in time domain and frequency domain...
详细信息
the need for early detection of diabetes mellitus has led to the development of various intelligent systems using machinelearning and artificial intelligence for the recognition of the presence of the disease. Howeve...
详细信息
datamining and big data are today the world’s leading technology. these techniques deal with diabetes in the banking sector, health services, cyber-security, voting, insurance, the real state, etc. Diabetes is a con...
详细信息
People frequently use technological means to collect information from many sources, share it with others on social media, and communicate with friends on these platforms. Social media is therefore a crucial new tool f...
详细信息
Graph few-shot learning aims to predict well by training with very few labeled data. Meta learning has been the most popular solution for few-shot learning problem. However, transductive linear probing shows that fine...
详细信息
ISBN:
(数字)9798350355925
ISBN:
(纸本)9798350355932
Graph few-shot learning aims to predict well by training with very few labeled data. Meta learning has been the most popular solution for few-shot learning problem. However, transductive linear probing shows that fine-tuning a simple linear classification head after a pretrained graph neural networks can outperforms most of the sophisticated-designed graph meta learning algorithms. therefore, in the paper, we propose a meta transductive linear probing methods named Meta-TLP to incorporate the advantages of graph self-supervised and graph meta learning model. Specifically, the graph neural network is firstly pretrained with graph contrastive learning methods. then we design an unsupervised meta training task construction methods to require meta tasks without relying on labeled data. Finally, we meta training the linear classification head on the meta training tasks to learn to fast adopt to novel classes. Experiment results show that our model can perform better than TLP on three real world datasets.
Due to limited resources, non-profit organizations and mobile app developers face challenges in implementing task-specific ranking models. Being trained on a large corpus of data, with demonstrated remarkable human-li...
详细信息
ISBN:
(数字)9798350355925
ISBN:
(纸本)9798350355932
Due to limited resources, non-profit organizations and mobile app developers face challenges in implementing task-specific ranking models. Being trained on a large corpus of data, with demonstrated remarkable human-like text generative performance, ChatGPT has shown promising potential in solving tasks from a wide range of categories. this study centers its attention on assessing ChatGPT's performance for content ranking tasks, which may help provide new efficient decision-making solutions that have significant societal impacts. Specifically, structured prompts are developed to guide ChatGPT to rank a set of volunteer tasks for various volunteer profiles. the results are then evaluated with a two-step human labeling and consistency score. Our findings demonstrate that ChatGPT's ranking preferences are highly consistent withthe human reviewer implying their effectiveness as efficiently accessible ranking models.
In view of the lack of personalization and interactivity in current oral English learning, a dialogue system based on reinforcement learning is introduced, aiming to improve users' oral expression ability by dynam...
详细信息
In view of the lack of personalization and interactivity in current oral English learning, a dialogue system based on reinforcement learning is introduced, aiming to improve users' oral expression ability by dynamically adjusting the learning path and feedback mechanism. the system adopts a deep Q-learning algorithm, optimizes the voice interaction effect through a reward mechanism, and realizes real-time adaptive adjustment of the dialogue. First, a dialogue framework based on reinforcement learning is designed, in which the user's voice input is converted into text and used as the input of the system. then, the system uses a deep Q-learning algorithm to conduct feedback learning based on the user's voice performance and grammatical errors, and adjusts the dialogue strategy and vocabulary recommendation in real time to improve the interactivity and accuracy of learning. Finally, the system trains multiple rounds of dialogues in a simulated environment to continuously optimize the speech recognition and dialogue response strategies. the whole process uses a reward mechanism to adjust the system behavior based on the actual performance of the user to ensure gradual improvement in the learning process. Learners with a medium foundation (intermediate) have a slight improvement in satisfaction scores, reaching 4.3 points, and their oral test improvement is +15 points. Learners with a high foundation (advanced) have a satisfaction score of 4.5 points in the use of the system, and their oral test improvement is +18 points, thanks to the detailed and accurate feedback provided by the system. By introducing a reinforcement learning algorithm, the designed English oral online dialogue system can adjust learning strategies in real time according to user feedback and improve learning effects. Future research will further optimize the reward mechanism and enhance the system's adaptive ability and intelligent recommendation function to better serve the needs of different learner
At present, the task of traffic flow forecasting poses significant challenges. Existing studies predominantly utilize spatial characteristics for comprehensive forecasting and do not extensively consider the issue of ...
详细信息
ISBN:
(数字)9798350355925
ISBN:
(纸本)9798350355932
At present, the task of traffic flow forecasting poses significant challenges. Existing studies predominantly utilize spatial characteristics for comprehensive forecasting and do not extensively consider the issue of extracting features from individual time series data. In the absence of known road topological structures, extracting effective temporal features of a single node becomes especially important. this paper proposes a new time series-based traffic prediction model-CNN-BiGRU-Attention(CBGA), which combines Convolutional Neural Networks (CNN), Bidirectional Gated Recurrent Units (BiGRU), and an attention mechanism. the model captures local features through CNN, captures long-term temporal dependencies through BiGRU, and introduces an attention mechanism to focus on key time steps, thereby effectively improving the forecasting performance. We conducted experiments using this model and compared it with traditional and modern models to demonstrate that the CBGA model has good performance in time series forecasting tasks and can effectively perform time series-based traffic flow forecasting tasks.
Most existing methods for operational intent recognition of air targets rely on complex models, leading to difficulties in accurately and efficiently identifying the true intent of the target. To overcome this challen...
详细信息
ISBN:
(数字)9798331530334
ISBN:
(纸本)9798331530341
Most existing methods for operational intent recognition of air targets rely on complex models, leading to difficulties in accurately and efficiently identifying the true intent of the target. To overcome this challenge, this paper constructs an air target intent feature set by carefully selecting relevant features and encoding them into time-series data. Furthermore, a novel deep learning approach based on the Transformer model is proposed, aiming to enhance the capability of operational intent recognition for air targets. the Transformer model not only excels in extracting potential feature information from the data but also effectively captures the global relationships within the battlefield situation data information. Additionally, through its self-attention mechanism, Transformer adeptly handles long-term dependencies within sequences. Experimental results demonstrate that the Transformer model achieves an accuracy of 98.86% in air target intent recognition tasks, highlighting its superior performance in this domain.
A computational pragmatic model of conversational scalar implicature is established by using Bayes' theorem and its statistical methods, and a small man-machine dialogue program is implemented based on it. Firstly...
详细信息
ISBN:
(数字)9798350355925
ISBN:
(纸本)9798350355932
A computational pragmatic model of conversational scalar implicature is established by using Bayes' theorem and its statistical methods, and a small man-machine dialogue program is implemented based on it. Firstly, four ideas that Bayesian methods can be used in computational pragmatics are extracted, and their fit with scalar implicature is demonstrated. then, by determining state set, prior and conditional probability, the corresponding conversational implicature is obtained after calculating the posterior probability. Finally, based on this process, a man-machine dialogue program for the probability analysis of implicature is programmed. this research shows that the conversational implicature module in natural language processing can be implemented by Bayesian-based causal inference.
暂无评论