Model Compression

This refers to techniques used to reduce the size and complexity of machine learning models, making them more efficient for deployment on resource-constrained devices. Methods include pruning unimportant connections, quantization to fewer bits, and knowledge distillation where a smaller model learns from a larger one.