ANS Data Encoding Without Tables or Division Overhead
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing asymmetric numeral system (ANS)-based encoding/decoding methods are memory-intensive and computationally costly due to division, modulo operations, and large encoding/decoding tables, particularly in AI compression scenarios with extensive distributions.
Innovation Solution
A data encoding method that utilizes a first and second operation to obtain encoded data without division, reducing memory usage by eliminating the need for extensive tables, and a data decoding method that uses mapping relationships to minimize memory and delay.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If existing ANS-based encoding is used, then compression efficiency is improved, but memory usage increases due to large encoding/decoding tables
Solution Approach 1:
The patent extracts and removes the large encoding/decoding tables from the ANS-based encoding system. Instead of using extensive pre-computed tables, the invention implements a tableless approach where encoding operations are performed through direct mathematical computations, thereby reducing memory usage while preserving compression efficiency
Solution Approach 2:
The patent replaces the mechanical table-lookup system with a computational system based on mathematical operations. The encoding process substitutes table-based retrieval with arithmetic computations involving the current state and input symbol, eliminating the need for large stored tables while achieving the same encoding function
2Productivity
If existing ANS-based encoding is used, then compression efficiency is improved, but computational complexity increases due to division and modulo operations
Solution Approach 1:
The patent replaces complex division and modulo operations with simpler arithmetic operations. The new method uses multiplication, addition, and bit manipulation operations that are computationally less intensive while achieving the same encoding results, thereby reducing computational complexity
Solution Approach 2:
The patent changes the mathematical parameters and operations used in the encoding process. Instead of using division and modulo operations with potentially large numbers, the invention uses multiplication by pre-computed constants and bit shifting operations, which are faster and require less computational resources
3Adaptability or versatility
If extensive distribution scenarios like AI compression are used, then encoding capability is improved, but memory occupation increases due to large encoding/decoding tables
Solution Approach 1:
The patent extracts and removes the large encoding/decoding tables that are particularly problematic in extensive distribution scenarios like AI compression. The tableless design allows the system to handle diverse input distributions without requiring proportionally larger memory resources
Solution Approach 2:
The patent creates a universal encoding approach that works across different distribution scenarios without requiring scenario-specific tables. The mathematical operations are designed to be distribution-agnostic, allowing the same lightweight implementation to handle various encoding situations including AI compression workloads
Data Source
AI summary
A data decoding method includes obtaining status information of a decoder and a memory bit, where the status information corresponds to a current to-be-decoded bit; and determining, based on a first mapping relationship, first information, second information, and third information that correspond to the status information, where the first information is a target symbol corresponding to the first bit, the target symbol is a decoding result of the first bit, the second information indicates a quantity of bits selected from the memory bit, the third information performs a summation operation with a bit selected from the memory bit based on the third second information, a result of the summation operation updates the status information, updated status information corresponds to a second bit, and the second bit is a to-be-decoded bit after the first bit.