grydra 1.0.0

grydra is a Ruby library for creating, training, and experimenting with neural networks. It includes neuron and layer classes, multiple modern activation functions (Tanh, ReLU, Leaky ReLU, Sigmoid, Swish, GELU), regularization techniques (Dropout, L1, L2), weight initialization (Xavier, He), data normalization (Z-Score, Min-Max), synthetic data generation, and simplified PCA. It offers evaluation metrics (MSE, MAE, accuracy, F1, confusion matrix, AUC-ROC), a configurable Adam optimizer, advanced training features like mini-batch, early stopping, and learning rate decay, as well as cross-validation and hyperparameter search. Furthermore, it allows visualizing architectures in ASCII or exporting them to Graphviz, working with text processing (binary vectorization and TF-IDF), and saving/loading models along with their vocabulary. It includes a simplified interface (EasyNetwork) for quickly training with numerical data, hashes, or text.