Problem
Research questions and friction points this paper is trying to address.
Source Coding
Coding for Machines
Task Scope
Information Retention
Feature Coding
On Task Scope and Information Retention in Source Coding
🤖 AI Summary
This study addresses the limitation of traditional source coding, which is typically partitioned by receiver type (human or machine) while overlooking the fundamental influence of task scope on information retention requirements. To overcome this, we propose a unified coding framework grounded in task scope, employing information-theoretic derivations and theoretical analysis of finite task families to systematically investigate rate relationships between source and feature coding under encoder observation constraints. We demonstrate that the minimum coding rate is determined by task scope rather than receiver identity, revealing that machine tasks may necessitate higher rates. Furthermore, we show that as task scope expands, the rates of these two coding paradigms diverge, thereby dismantling the conventional human-machine coding dichotomy. This work redefines the criteria for information discarding in source coding, establishing task scope as the governing principle for determining information retention.
📝 Abstract
We argue that dividing codec design into Coding for Machines (CfM) and Coding for Humans (CfH) is a misleading distinction for deciding what information a codec may discard. Receiver identity does not determine admissible information loss. The required rate depends on task scope, including the predictions to support, their losses and tolerated risks, the encoder observation, and the permitted decoding procedures. Notably, a machine task may have a higher minimum rate than a restricted human decision. Rate savings on selected machine tasks apply only to the stated requirements, not to an intrinsic ordering by receiver type. We extend source and feature coding to finite task families, derive when restricting the encoder observation preserves the minimum rate, and show that equality between source and split-feature coding rates can no longer hold as the task scope expands.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Source Coding
Coding for Machines
Task Scope
Feature Coding
Information Retention
🔎 Similar Papers
No similar papers found.
