Entropy Model

Entropy models are statistical tools used to estimate the probability distribution of data, primarily within the context of data compression and machine learning. Current research focuses on developing efficient entropy models, particularly for images and videos, often employing transformer-based architectures or convolutional neural networks with attention mechanisms to capture both local and global dependencies within data. These advancements are improving the performance of learned image and video compression, leading to higher compression ratios and better reconstruction quality, with applications ranging from multimedia codecs to resource-constrained IoT devices and federated learning scenarios.