Tiêu đề G12S2 Session 09.pdf ✅

G11S1 Session 09 (Explore) [Multi-Class Classification] (5 Minutes) Multi-class classification: - classifying instances into one of three or more classes - such as face recognition, plant species identification, or optical character recognition. - Unlike binary classification, there is no distinction between normal and abnormal outcomes—each instance belongs to one of several known classes. - Some problems involve a large number of possible class labels, like recognizing a face among thousands. Predicting word sequences, such as in text translation, is also a form of multi-class classification, where each word belongs to one of many possible classes, defined by the vocabulary size. (Explore) [CNN Architecture Multi-Class Classification] (5 Minutes) Multi-class classification trains a model to categorize data into one of several predefined classes. CNNs (Convolutional Neural Networks) are ideal for image tasks, automatically learning features from input data. Here's a brief explanation: 1. Input Layer: Takes raw data, often images.
2. Convolutional Layers: Extract features by applying filters to detect patterns like edges and textures. 3. Activation Function (e.g., ReLU): o Introduces non-linearity to help learn complex relationships. 4. Pooling Layers: o Down-sample || Reduce the spatial dimensions of the feature maps, lowering computation and improving robustness to variations in object position and scale. 5. Flatten Layer: o Flattens the high-dimensional feature maps into a one-dimensional vector o بنضرب كل الـ Dimensions في بعض o preparing the data for fully connected layers. 6. Fully Connected Layers: o Dense layers that combine features learned by convolutional layers for making predictions. (Hidden Layers) o Last fully connected layer has nodes equal to the number of classes. (Output layer) 7. Output Layer: o Produces a probability distribution across all classes using the softmax activation function. || In binary classification, we used the sigmoid function. o The class with the highest probability is chosen as the predicted class for a given input. The softmax function: converts a vector of real numbers into a probability distribution, often used in neural networks for multi-class classification. - It enables the model to predict the most likely class while providing confidence levels for each class through probabilities. - Paired with categorical cross-entropy loss, softmax is ideal for optimizing models in multi-class classification tasks. 8. Loss Function: o Categorical Cross-Entropy is commonly used as the loss function for multi-class classification in CNNs. o The difference between the predicted probability and the actual class labels. 9. Optimizer (e.g., SGD, Adam): o Adjusts the weights and biases based on the gradients of the loss function during training. 10. Training Data: o The model is trained on a labeled dataset containing examples of each class. 11. Validation and Test Data: o Separate datasets are used to evaluate the model's performance during training (validation) and after training (test). 12. Data Augmentation (Optional): o Enhances training with transformations (e.g., rotations, flips) to improve generalization. o طريقة نكبر بيها ال dataset بردو

Visualized the architecture using visualkeras.