Epoch

An epoch is the cycle in machine learning and deep learning in which the entire training dataset is passed through the model once. One epoch means that all samples in the training data have been used once to update the model's weights. In practice, a single epoch is rarely sufficient for the model to learn well, so training is repeated over multiple epochs to improve accuracy. Key terms and concepts related to epochs: • Number of epochs: How many times the training data is fed through the model in total • Batch size: The number of data samples used in one training step; the full dataset is divided into multiple batches • Iteration: The number of training steps per batch. 1 epoch = (total number of samples ÷ batch size) iterations For example, with 10,000 training samples and a batch size of 100, one epoch consists of 100 iterations. Too few epochs leads to underfitting, while too many can cause overfitting, so choosing the right number of epochs is critical to model performance. Common practice is to monitor how the validation loss and accuracy change over time and select the optimal number of epochs accordingly. A technique called "early stopping" is also widely used, where training is automatically halted when validation accuracy stops improving for a set number of epochs. In short, an epoch is an important metric showing "how many times the AI model has learned from the training data," and is a key concept for adjusting the quality and quantity of training.