Quantized Entropy Coding Parameters for Neural Media Compression

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Neural network variables related to entropy coding, when quantized, cause significant degradation in compression efficiency and cannot be recovered by re-training, especially in common use cases, and the specific properties of these variables are not adequately addressed by general quantization tools.

Innovation Solution

Optimizing the definition of entropy coding variables by determining a probability distribution function parameter based on a distribution optimized for quantization, using empirical methods or algorithms that evaluate coding redundancy or solve ordinary differential equations, to preserve important information and minimize memory usage.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If neural network variables are quantized to low-precision integers to improve speed and reduce power consumption, then processing efficiency is improved, but compression efficiency deteriorates due to information loss in entropy coding variables

Engineering Contradiction:

Improveprocessing speedVSAvoidcompression efficiency

Core Design Contradiction:

ProductivityVSLoss of information

Solution Approach 1:

The patent transforms the probability distribution function parameter through a monotonic function to create a new parameter that is optimized for quantization. This parameter transformation preserves the essential information needed for entropy coding while making the variable more suitable for low-precision representation, thereby resolving the contradiction between quantization efficiency and compression performance

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent applies a monotonic transformation function to the probability distribution parameter before quantization occurs. This preliminary action ensures that the quantized version retains the necessary statistical properties for effective entropy coding, preventing information loss that would otherwise occur with direct quantization

Inventive Principle:

Principle #10Preliminary action

2Ease of manufacture

If general quantization tools are used on entropy coding variables, then implementation is simplified, but compression performance deteriorates because specific properties of entropy coding variables are not accounted for

Engineering Contradiction:

Improveimplementation simplicityVSAvoidcompression performance

Core Design Contradiction:

Ease of manufactureVSLoss of information

Solution Approach 1:

The patent applies a specific monotonic transformation tailored to the properties of probability distribution parameters used in entropy coding. This localized transformation approach accounts for the specific statistical characteristics of these variables, ensuring optimal performance while maintaining implementation feasibility through a systematic transformation process

Inventive Principle:

Principle #3Local quality

3Loss of information

If high-precision floating-point arithmetic is used for neural network entropy coding, then compression efficiency is maintained, but power consumption and computational complexity increase

Engineering Contradiction:

Improvecompression efficiencyVSAvoidpower consumption

Core Design Contradiction:

Loss of informationVSUse of energy by moving object

Solution Approach 1:

The patent transforms the probability distribution parameter using a monotonic function to create a quantization-friendly representation. This parameter change enables the use of low-precision integer arithmetic while preserving the statistical information necessary for effective entropy coding, thereby reducing power consumption without sacrificing compression efficiency

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent replaces expensive high-precision floating-point operations with cheaper low-precision integer operations through parameter transformation. The transformed parameter allows entropy coding to be performed efficiently with simplified arithmetic, reducing computational cost and power consumption while maintaining adequate compression performance

Inventive Principle:

Principle #27Cheap short-living objects (Disposable)

Data Source

PatentUS12470715B2Neural-network media compression using quantized entropy coding distribution parameters

Publication Date: 2025.11.11 QUALCOMM INC

AI summary

A media coder performs entropy coding techniques for media data coded using neural-based techniques. A media coder is configured to determine a probability distribution function parameter for a data element of a data stream coded by a neural-based media compression technique, wherein the probability distribution function parameter is a function of a standard deviation of a probability distribution function of the data stream, determine a code vector based on the probability distribution function parameter, and entropy code the data element using the code vector.