Titre :	Generative models based on GANs and RNNs for phobia data generation and classification
Type de document :	texte imprimé
Auteurs :	Abdeldjalil Hani, Auteur ; Mohamed Said Benabdallah ; Djemame ,Safia, Directeur de thèse
Editeur :	Setif:UFA
Année de publication :	2024
Importance :	1 vol (76 f .)
Format :	29 cm
Langues :	Anglais (eng)
Catégories :	Thèses & Mémoires:Informatique
Mots-clés :	Phobia Detection Classification Augmentation AI conversational models GANs RNNs Rare data
Index. décimale :	004 - Informatique
Résumé :	Phobia detection and classification pose significant challenges due to the rarity of comprehensive datasets encompassing both visual and textual data. For the case of textual data, doctors rarely record patient oral description of the phobia triggering factors posing the problem of lack of textual descriptive data that portray the real cause of fear. This study leverages large AI Conversational models to augment and classify phobia-related data effectively where the first batch of data is generated using conversational models. GANs are utilized for the generation of high-quality synthetic images that represent various phobia-inducing scenarios, enhancing the diversity and volume of visual data. Concurrently, RNNs are employed to augment textual descriptive data by generating realistic narratives and descriptions associated with different phobias, capturing the sequential and contextual features essential for accurate analysis and diagnosis. Each type of augmented data images and text is assessed independently by training classification models tailored for their respective data modalities. The obtained results of both data generation and phobia prediction exhibit a promissing performance of generative and AI based models in rare data augmentation as well as the performance of these data to train predictive models is remarkably interesting. In which we achieved an accuracy score up to 97% in distinguish five different phobias.
Note de contenu :	Sommaire General Introduction 1 Problematic and Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 Research Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 Thesis structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 1 Deep Learning and Generative Adversarial Networks 3 1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 1.2 Deep Learning (DL) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 1.2.1 Deep Neural Networks . . . . . . . . . . . . . . . . . . . . . . . . 3 1.2.2 Deep Neural Networks machanism . . . . . . . . . . . . . . . . . . 4 1.2.3 Convolutional Neural Networks . . . . . . . . . . . . . . . . . . . 5 1.2.4 Recurrent Neural Networks . . . . . . . . . . . . . . . . . . . . . 7 1.2.5 Loss Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 1.2.6 Training dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . 11 1.2.7 Activation Functions . . . . . . . . . . . . . . . . . . . . . . . . . 12 1.3 Exploring Generative Models and GANs . . . . . . . . . . . . . . . . . . 13 1.3.1 Generative Modeling . . . . . . . . . . . . . . . . . . . . . . . . . 13 1.3.2 Generative Adversarial Networks (GANs) . . . . . . . . . . . . . . 14 1.3.3 Key components: Generator and Discriminator . . . . . . . . . . 15 1.3.4 GANs mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 1.3.5 Math and Theory behind GANs . . . . . . . . . . . . . . . . . . . 16 1.3.6 Applications of GANs . . . . . . . . . . . . . . . . . . . . . . . . 19 1.3.7 Challenges in GANs . . . . . . . . . . . . . . . . . . . . . . . . . 19 1.3.8 Deep Convolutional GANs (DCGANs) . . . . . . . . . . . . . . . 21 1.3.9 Wasserstein GANs (WGANs) . . . . . . . . . . . . . . . . . . . . 22 1.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 2 Generative models and phobia diagnosis and treatment 24 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 2.2 Introduction to Phobias . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 2.2.1 Definition and Understanding of Phobias . . . . . . . . . . . . . . 24 2.2.2 Types of Phobias . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 2.2.3 Individuals who can suffer from Phobias . . . . . . . . . . . . . . 26 2.2.4 Causes of Phobias . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 2.2.5 The importance of treating Phobias . . . . . . . . . . . . . . . . . 26 2.3 Traditional treatement approaches . . . . . . . . . . . . . . . . . . . . . . 27 2.3.1 Overview of cognitive-behavioral therapy - CBT . . . . . . . . . . 27 2.3.2 Exposure therapy fundamentals . . . . . . . . . . . . . . . . . . . 28 2.3.3 The Efficacy of Exposure Therapy in Treating Fear . . . . . . . . 29 2.3.4 The Efficacy of Exposure versus Cognitive therapy . . . . . . . . 30 2.4 Computer Science Techniques used to fight Phobias . . . . . . . . . . . . 31 2.4.1 The Intersection of Computer Science and Mental Health . . . . . 31 2.4.2 Augmented Reality (AR) Applications in Exposure Therapy . . . 32 2.4.3 Virtual Reality (VR) Therapy: A Game-Changer in Phobia Treatment. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 2.4.4 Machine Learning and Phobia Detection: Innovations in Diagnosis 34 2.4.5 Benifits and Ethical Considerations in Technological Phobia Interventions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 2.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 3 AI and Generative models for phobia data 36 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 3.2 Contribution 1: using AI for data collection . . . . . . . . . . . . . . . . 37 3.2.1 ’Descriptions’ data-set . . . . . . . . . . . . . . . . . . . . . . . . 37 3.2.2 ’Images-Descriptions’ pairs data-set . . . . . . . . . . . . . . . . . 41 3.2.3 Advantages of using ChatGPT and Gemini . . . . . . . . . . . . . 46 3.3 Contribution 2: Data Augmentation Using GAN and RNN . . . . . . . . 47 3.3.1 GANs for images generation . . . . . . . . . . . . . . . . . . . . . 48 3.3.2 RNN for text generation . . . . . . . . . . . . . . . . . . . . . . . 52 3.4 General discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 3.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 4 Implementation and experimental results 54 4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54 4.2 Experimentation Environment . . . . . . . . . . . . . . . . . . . . . . . . 54 4.2.1 Google Drive . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54 4.2.2 Google Colab Pro . . . . . . . . . . . . . . . . . . . . . . . . . . . 54 4.2.3 PyTorch and TensorFlow . . . . . . . . . . . . . . . . . . . . . . . 54 4.2.4 Additional Libraries and Resources: . . . . . . . . . . . . . . . . . 55 4.3 AI conversational data generation results . . . . . . . . . . . . . . . . . . 55 4.4 GANs for images generation results . . . . . . . . . . . . . . . . . . . . . 58 4.5 RNN for text generation results . . . . . . . . . . . . . . . . . . . . . . . 61 4.6 Classification for phobia detection results . . . . . . . . . . . . . . . . . . 63 4.6.1 Data-set validation - Phobia detection . . . . . . . . . . . . . . . 63 4.6.2 Models used for text classification . . . . . . . . . . . . . . . . . . 63 4.6.3 Model used for image classification (CNN) . . . . . . . . . . . . . 65 4.6.4 Text classification . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 4.6.5 Image classification . . . . . . . . . . . . . . . . . . . . . . . . . . 67 4.7 General discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70 4.8 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70 General Conclusion 71
Côte titre :	MAI/0905

Generative models based on GANs and RNNs for phobia data generation and classification [texte imprimé] / Abdeldjalil Hani, Auteur ; Mohamed Said Benabdallah ; Djemame ,Safia, Directeur de thèse . - [S.l.] : Setif:UFA, 2024 . - 1 vol (76 f .) ; 29 cm.
Langues : Anglais (eng)

Catégories :	Thèses & Mémoires:Informatique
Mots-clés :	Phobia Detection Classification Augmentation AI conversational models GANs RNNs Rare data
Index. décimale :	004 - Informatique
Résumé :	Phobia detection and classification pose significant challenges due to the rarity of comprehensive datasets encompassing both visual and textual data. For the case of textual data, doctors rarely record patient oral description of the phobia triggering factors posing the problem of lack of textual descriptive data that portray the real cause of fear. This study leverages large AI Conversational models to augment and classify phobia-related data effectively where the first batch of data is generated using conversational models. GANs are utilized for the generation of high-quality synthetic images that represent various phobia-inducing scenarios, enhancing the diversity and volume of visual data. Concurrently, RNNs are employed to augment textual descriptive data by generating realistic narratives and descriptions associated with different phobias, capturing the sequential and contextual features essential for accurate analysis and diagnosis. Each type of augmented data images and text is assessed independently by training classification models tailored for their respective data modalities. The obtained results of both data generation and phobia prediction exhibit a promissing performance of generative and AI based models in rare data augmentation as well as the performance of these data to train predictive models is remarkably interesting. In which we achieved an accuracy score up to 97% in distinguish five different phobias.
Note de contenu :	Sommaire General Introduction 1 Problematic and Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 Research Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 Thesis structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 1 Deep Learning and Generative Adversarial Networks 3 1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 1.2 Deep Learning (DL) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 1.2.1 Deep Neural Networks . . . . . . . . . . . . . . . . . . . . . . . . 3 1.2.2 Deep Neural Networks machanism . . . . . . . . . . . . . . . . . . 4 1.2.3 Convolutional Neural Networks . . . . . . . . . . . . . . . . . . . 5 1.2.4 Recurrent Neural Networks . . . . . . . . . . . . . . . . . . . . . 7 1.2.5 Loss Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 1.2.6 Training dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . 11 1.2.7 Activation Functions . . . . . . . . . . . . . . . . . . . . . . . . . 12 1.3 Exploring Generative Models and GANs . . . . . . . . . . . . . . . . . . 13 1.3.1 Generative Modeling . . . . . . . . . . . . . . . . . . . . . . . . . 13 1.3.2 Generative Adversarial Networks (GANs) . . . . . . . . . . . . . . 14 1.3.3 Key components: Generator and Discriminator . . . . . . . . . . 15 1.3.4 GANs mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 1.3.5 Math and Theory behind GANs . . . . . . . . . . . . . . . . . . . 16 1.3.6 Applications of GANs . . . . . . . . . . . . . . . . . . . . . . . . 19 1.3.7 Challenges in GANs . . . . . . . . . . . . . . . . . . . . . . . . . 19 1.3.8 Deep Convolutional GANs (DCGANs) . . . . . . . . . . . . . . . 21 1.3.9 Wasserstein GANs (WGANs) . . . . . . . . . . . . . . . . . . . . 22 1.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 2 Generative models and phobia diagnosis and treatment 24 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 2.2 Introduction to Phobias . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 2.2.1 Definition and Understanding of Phobias . . . . . . . . . . . . . . 24 2.2.2 Types of Phobias . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 2.2.3 Individuals who can suffer from Phobias . . . . . . . . . . . . . . 26 2.2.4 Causes of Phobias . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 2.2.5 The importance of treating Phobias . . . . . . . . . . . . . . . . . 26 2.3 Traditional treatement approaches . . . . . . . . . . . . . . . . . . . . . . 27 2.3.1 Overview of cognitive-behavioral therapy - CBT . . . . . . . . . . 27 2.3.2 Exposure therapy fundamentals . . . . . . . . . . . . . . . . . . . 28 2.3.3 The Efficacy of Exposure Therapy in Treating Fear . . . . . . . . 29 2.3.4 The Efficacy of Exposure versus Cognitive therapy . . . . . . . . 30 2.4 Computer Science Techniques used to fight Phobias . . . . . . . . . . . . 31 2.4.1 The Intersection of Computer Science and Mental Health . . . . . 31 2.4.2 Augmented Reality (AR) Applications in Exposure Therapy . . . 32 2.4.3 Virtual Reality (VR) Therapy: A Game-Changer in Phobia Treatment. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 2.4.4 Machine Learning and Phobia Detection: Innovations in Diagnosis 34 2.4.5 Benifits and Ethical Considerations in Technological Phobia Interventions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 2.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 3 AI and Generative models for phobia data 36 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 3.2 Contribution 1: using AI for data collection . . . . . . . . . . . . . . . . 37 3.2.1 ’Descriptions’ data-set . . . . . . . . . . . . . . . . . . . . . . . . 37 3.2.2 ’Images-Descriptions’ pairs data-set . . . . . . . . . . . . . . . . . 41 3.2.3 Advantages of using ChatGPT and Gemini . . . . . . . . . . . . . 46 3.3 Contribution 2: Data Augmentation Using GAN and RNN . . . . . . . . 47 3.3.1 GANs for images generation . . . . . . . . . . . . . . . . . . . . . 48 3.3.2 RNN for text generation . . . . . . . . . . . . . . . . . . . . . . . 52 3.4 General discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 3.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 4 Implementation and experimental results 54 4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54 4.2 Experimentation Environment . . . . . . . . . . . . . . . . . . . . . . . . 54 4.2.1 Google Drive . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54 4.2.2 Google Colab Pro . . . . . . . . . . . . . . . . . . . . . . . . . . . 54 4.2.3 PyTorch and TensorFlow . . . . . . . . . . . . . . . . . . . . . . . 54 4.2.4 Additional Libraries and Resources: . . . . . . . . . . . . . . . . . 55 4.3 AI conversational data generation results . . . . . . . . . . . . . . . . . . 55 4.4 GANs for images generation results . . . . . . . . . . . . . . . . . . . . . 58 4.5 RNN for text generation results . . . . . . . . . . . . . . . . . . . . . . . 61 4.6 Classification for phobia detection results . . . . . . . . . . . . . . . . . . 63 4.6.1 Data-set validation - Phobia detection . . . . . . . . . . . . . . . 63 4.6.2 Models used for text classification . . . . . . . . . . . . . . . . . . 63 4.6.3 Model used for image classification (CNN) . . . . . . . . . . . . . 65 4.6.4 Text classification . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 4.6.5 Image classification . . . . . . . . . . . . . . . . . . . . . . . . . . 67 4.7 General discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70 4.8 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70 General Conclusion 71
Côte titre :	MAI/0905