Rate-distortion optimization for transformer inference
Split computing for language models, extending the theory of usable information.
Split Computing
Lossy common information in a learnable Gray-Wyner Network
Isolate the common information between two dependent computer vision tasks.
Rate-distortion theory in coding for machines and its applications
Theoretical considerations and evaluation of split and distillation points.
Towards task-compatible compressible representations
Task reconstruction loss acts as a regularizer, increasing rate-distortion performance in coding for humans and machines.
Base layer efficiency in scalable human-machine coding
Improving the shared channel in coding for machines (CfM).
Conditional and residual methods in scalable coding for humans and machines
A comparison between conditional and residual entropy codecs for a two-channel systems of tasks with nested information.