Genomic Arithmetic Coding With ML Predictors and Dynamic Contexts
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The current MPEG-G standard for genomic data compression is limited by its restrictive context selection, lack of support for advanced predictors like neural networks, and absence of multi-symbol arithmetic coding, which hampers efficient compression and adaptability.
Innovation Solution
An extensible framework is introduced that allows for the selection of multiple contexts, use of machine learning models such as neural networks for prediction, and support for static, semi-adaptive, and adaptive training modes, along with multi-symbol arithmetic coding, enabling improved compression by capturing complex correlations and dependencies in genomic data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If context-adaptive binary arithmetic coding (CABAC) is used for compression, then compression efficiency is improved, but the system is limited by restrictive context selection and lack of support for advanced predictors
Solution Approach 1:
The patent implements dynamic context selection where the context for arithmetic coding is not fixed but can be adaptively chosen from multiple sources including previous symbols, other descriptors, and external data. The system dynamically adjusts context based on the specific data being encoded, allowing flexible adaptation to different genomic data patterns while maintaining compression efficiency.
Solution Approach 2:
The framework provides a universal context selection mechanism that can work with multiple types of data sources and predictor types (including advanced machine learning predictors). The same arithmetic coding infrastructure supports both traditional CABAC and more advanced prediction methods, making the system versatile across different genomic data types and compression requirements.
2Ease of operation
If only previous symbols are used as context in CABAC, then the decoding process is simple, but complex correlations in genomic data cannot be captured
Solution Approach 1:
The patent extends the context dimension beyond just previous symbols by incorporating additional data sources such as other descriptors, genomic position information, and external reference data. This multi-dimensional context approach captures complex correlations in genomic data while maintaining a systematic decoding process that builds upon the simple CABAC foundation.
3Measurement precision
If advanced machine learning predictors are introduced, then prediction accuracy is improved, but computational complexity increases
Solution Approach 1:
The system dynamically selects between different predictor types including both simple statistical predictors and advanced machine learning models based on the specific data characteristics and computational requirements. This dynamic selection allows the system to achieve high prediction accuracy when needed while falling back to simpler methods when computational resources are constrained.
Solution Approach 2:
The framework allows adjustment of predictor complexity as a configurable parameter. Users can select from multiple predictor types with varying computational requirements, and the system can adapt the predictor complexity based on the specific genomic data being processed, available computational resources, and desired compression performance.
4Productivity
If multiple contexts and advanced predictors are supported, then compression ratio is improved, but implementation complexity increases
Solution Approach 1:
The patent segments the context selection and prediction process into distinct, modular components. Each context source and predictor type is implemented as a separate module that can be independently configured and selected. This segmentation allows the system to achieve high compression ratios through multiple contexts and advanced predictors while managing implementation complexity through modular design and clear separation of concerns.
Data Source
AI summary
A method and a system for decoding MPEG-G encoded data of genomic information, including: receiving MPEG-G encoded data; extracting encoding parameters; selecting an arithmetic decoding type based upon the extracted encoding parameters; selecting a predictor type specifying the method to obtain probabilities of symbols which were used for arithmetically encoding the data, based upon the extracted encoding parameters; selecting arithmetic coding contexts based upon the extracted encoding parameters; and decoding the encoded data using the selected predictor and the selected arithmetic coding contexts.