Table 1 Computational resources, frameworks, and training configuration.
From: Towards improved fake news detection using a hybrid RoBERTa and metadata enhanced XGBoost model
Component | Details |
|---|---|
Computing hardware | Intel Core i9-12900K (16C/24T), NVIDIA RTX 4090 (24GB VRAM), 64GB DDR5 RAM |
Operating system | Windows 11 Pro |
Deep learning frameworks | PyTorch 2.0, TensorFlow 2.10, Keras 2.11 |
GPU acceleration | CUDA 12.1, cuDNN 8.9 |
Datasets used | PolitiFact, GossipCop |
Preprocessing | BERT-based Tokenization, stopword removal, lemmatization |
Model architecture | Transformer-based Embeddings + TF-IDF Features + XGBoost Classifier |
Training duration | 100 Epochs, Batch Size = 16 |
Learning rate adjustment | ReduceLROnPlateau (Triggers after 3 epochs of no improvement) |
Regularization techniques | Early stopping (patience: 10 epochs), gradient clipping |
Evaluation metrics | Accuracy, precision, recall, F1-score |
