Recurrent Neural Networks (RNNs): Architectures, Training Tricks, and Introduction to Influential Research

Neuromethods, Journal Year: 2023, Volume and Issue: unknown, P. 117 - 138

Published: Jan. 1, 2023

Abstract Recurrent neural networks (RNNs) are network architectures with hidden state and which use feedback loops to process a sequence of data that ultimately informs the final output. Therefore, RNN models can recognize sequential characteristics in help predict next likely point sequence. Leveraging power processing, cases tend be connected either language or time-series analysis. However, multiple popular have been introduced field, starting from SimpleRNN LSTM deep RNN, applied different experimental settings. In this chapter, we will present six distinct highlight pros cons each model. Afterward, discuss real-life tips tricks for training models. Finally, four modeling applications –text classification, summarization, machine translation, image-to-text translation– thereby demonstrating influential research field.

Language: Английский

---

Artificial Intelligence for Cognitive Health Assessment: State-of-the-Art, Open Challenges and Future Directions

Cognitive Computation, Journal Year: 2023, Volume and Issue: 15(6), P. 1767 - 1812

Published: June 24, 2023

Abstract The subjectivity and inaccuracy of in-clinic Cognitive Health Assessments (CHA) have led many researchers to explore ways automate the process make it more objective facilitate needs healthcare industry. Artificial Intelligence (AI) machine learning (ML) emerged as most promising approaches CHA process. In this paper, we background delve into extensive research recently undertaken in domain provide a comprehensive survey state-of-the-art. particular, careful selection significant works published literature is reviewed elaborate range enabling technologies AI/ML techniques used for CHA, including conventional supervised unsupervised learning, deep reinforcement natural language processing, image processing techniques. Furthermore, an overview various means data acquisition benchmark datasets. Finally, discuss open issues challenges using AI ML along with some possible solutions. summary, paper presents tools, lists methods provides technological advancements, usage issues, domain. We hope first-of-its-kind will significantly contribute identifying gaps complex rapidly evolving interdisciplinary mental health field.

Language: Английский

---

State-of-the-Art of Stress Prediction from Heart Rate Variability Using Artificial Intelligence

Cognitive Computation, Journal Year: 2023, Volume and Issue: 16(2), P. 455 - 481

Published: Oct. 12, 2023

Abstract Recent advancements in the manufacturing and commercialisation of miniaturised sensors low-cost wearables have enabled an effortless monitoring lifestyle by detecting analysing physiological signals. Heart rate variability (HRV) denotes time interval between consecutive heartbeats.The HRV signal, as detected devices, has been popularly used indicative measure to estimate level stress, depression, anxiety. For years, artificial intelligence (AI)-based learning systems known for their predictive capabilities, recent AI models with deep (DL) architectures successfully applied achieve unprecedented accuracy. In order determine effective methodologies collection, processing, prediction stress from data, this work presents depth analysis 43 studies reporting application various algorithms. The methods are summarised tables thoroughly evaluated ensure completeness findings reported results. To make comprehensive, a detailed review conducted on sensing technologies, pre-processing multi-modal employed models. This is followed critical examination how Machine Learning (ML) models, utilised predicting data. addition, reseults selected carefully analysed identify features that enable perform better. Finally, challenges using predict listed, along some possible mitigation strategies. aims highlight impact AI-based expected aid development more meticulous techniques.

Language: Английский

---

Enhancing biofeedback-driven self-guided virtual reality exposure therapy through arousal detection from multimodal data using machine learning

Brain Informatics, Journal Year: 2023, Volume and Issue: 10(1)

Published: June 21, 2023

Virtual reality exposure therapy (VRET) is a novel intervention technique that allows individuals to experience anxiety-evoking stimuli in safe environment, recognise specific triggers and gradually increase their perceived threats. Public-speaking anxiety (PSA) prevalent form of social anxiety, characterised by stressful arousal generated when presenting an audience. In self-guided VRET, participants can tolerance reduce anxiety-induced PSA over time. However, creating such VR environment determining physiological indices or distress open challenge. Environment modelling, character creation animation, psychological state determination the use machine learning (ML) models for stress detection are equally important, multi-disciplinary expertise required. this work, we have explored series ML with publicly available data sets (using electroencephalogram heart rate variability) predict states. If detect arousal, trigger calming activities allow cope overcome distress. Here, discuss means effective selection parameters detection. We propose pipeline model problem different parameter settings context virtual therapy. This be extended other domains interest where crucial. Finally, implemented biofeedback framework VRET successfully provided feedback as brain laterality index from our acquired multimodal anxiety.

Language: Английский

---

Cross-Content Recommendation between Movie and Book using Machine Learning

Published: Oct. 13, 2021

Machine learning-driven recommendation systems are widely used in today's growing digital world. Existing movie and book recommender work using a collaborative approach, which can result lack of fresh diverse content reduced surprise factor. There is also no platform providing recommendations across different contents, such as for books from movies vice versa. In this paper, our main goal to introduce cross-content system based on the descriptions identifying similarities natural language processing machine learning algorithms. We processed combined dataset two types generated TF-IDF vector apply three algorithms: K-means clustering, hierarchical cosine similarity. being known research similar with ground truth labels, we applied subjective reasoning evaluate results system.

Language: Английский

---

Evaluation of an Arabic Chatbot Based on Extractive Question-Answering Transfer Learning and Language Transformers

AI, Journal Year: 2023, Volume and Issue: 4(3), P. 667 - 691

Published: Aug. 16, 2023

Chatbots are programs with the ability to understand and respond natural language in a way that is both informative engaging. This study explored current trends of using transformers transfer learning techniques on Arabic chatbots. The proposed methods used various semantic embedding models from AraBERT, CAMeLBERT, AraElectra-SQuAD, AraElectra (Generator/Discriminator). Two datasets were for evaluation: one 398 questions, other 1395 questions 365,568 documents sourced Wikipedia. Extensive experimental works conducted, evaluating manually crafted entire set by confidence similarity metrics. Our results demonstrate combining power transformer architecture extractive chatbots can provide more accurate contextually relevant answers Arabic. Specifically, our showed AraElectra-SQuAD model consistently outperformed models. It achieved an average score 0.6422 0.9773 first dataset, 0.6658 0.9660 second dataset. concludes remarkable performance, high confidence, robustness, which highlights its potential practical applications processing tasks suggests be further enhanced tasks, such as specialized chatbots, virtual assistants, information retrieval systems Arabic-speaking users.

Language: Английский

---

Analysis and Evaluation of the Impact of Integrating Mental Health Education into the Teaching of University Civics Courses in the Context of Artificial Intelligence

Wireless Communications and Mobile Computing, Journal Year: 2022, Volume and Issue: 2022, P. 1 - 11

Published: Aug. 2, 2022

In higher education teaching work, college students not only need to master the professional knowledge and skills they learn during their school study but also improve self-education self-cultivation constantly comprehensive ability of learning. At present, there are differences relationships between in civic mental normal education, how play role integration two educations has become a problem that needs be considered current work students’ training. The can make up for shortcomings monolithic optimize methods them from certain perspective achieve development goals complementing each other being independent other, so understand more learning contents promote own minds minds. response mind-normal cannot automatically integrated into university thought political science courses, context artificial intelligence, this paper proposes multi-channel-based ideological information fusion model. model channels, BERT+CNN BERT BiLSTM-Attention; firstly, pretraining is used obtain word vector representation fused text context; then, CNN network channel one enhance local feature extraction text, BiLSTM-Attention enhances long sequence processing key. Finally, features 1 2 classified using softmax excitation function. To verify effectiveness proposed model, experiments conducted on public datasets demonstrate method.

Language: Английский

---

Arabic Chatbot Evaluation Based on Extractive Question-Answering Transfer Learning and Language Transformers

Published: July 10, 2023

Chatbots are computer programs that use artificial intelligence to imitate human conversations. Recent advancements in deep learning have shown interest utilizing language transformers, which do not rely on predefined rules and responses like traditional chatbots. This study provides a comprehensive review of previous research chatbots employ transfer models. Specifically, it examines the current trends using transformers with techniques evaluate ability Arabic understand conversation context demonstrate natural behavior. The proposed methods explore AraBERT, CAMeLBERT, AraElectra-SQuAD, AraElectra (Generator/Discriminator) different variants these semantic embedding Two datasets were used for evaluation: one 398 questions corresponding documents, another 1395 365,568 documents sourced from Wikipedia. Extensive experimental works conducted, evaluating both manually crafted entire set questions, confidence similarity metrics. results showed AraElectra-SQuAD model achieved an average score 0.6422 0.9773 first dataset, 0.6658 0.9660 second dataset. concludes consistently outperformed other models, displaying remarkable performance, high confidence, scores, as well robustness, highlighting its potential practical applications processing tasks suggests can be further enhanced applied various such chatbots, virtual assistants, information retrieval systems Arabic-speaking users. By combining power transformer architecture fine-tuning SQuAD-like large data, this trend demonstrates provide accurate contextually relevant answers Arabic.

Language: Английский

---

Speech Emotion Recognition: An Empirical Analysis of Machine Learning Algorithms Across Diverse Data Sets

Communications in computer and information science, Journal Year: 2024, Volume and Issue: unknown, P. 32 - 46

Published: Jan. 1, 2024

Language: Английский

---

Brain Informatics

Lecture notes in computer science, Journal Year: 2021, Volume and Issue: unknown

Published: Jan. 1, 2021

Language: Английский

---