Deep Learning
Deep learning is a branch of machine learning based on multi-layer artificial neural networks, capable of automatic feature extraction from raw data.
Convolutional Neural Networks (CNN)
Designed for grid-structured image data, using convolution kernels to extract local spatial features, widely applied in image classification and detection.
Transformer Architecture
Proposed in 2017, relying on self-attention mechanism. Became foundation of modern LLMs, multi-modal models and computer vision large models.
Backpropagation
Core training algorithm, computes gradient loss layer by layer backward, updates network weights with gradient descent optimizer.