Neural Upsampling of Financial Time-Series After Lossy Compression

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current lossy compression techniques in data compression result in data loss, particularly in applications like video streaming and financial time-series data, where recovering lost information is challenging, especially in high-frequency or long-duration datasets.

Innovation Solution

A neural network system that incorporates a novel AI deblocking network with recurrent layers for feature extraction and a channel-wise transformer with attention to capture complex inter-channel dependencies, enabling effective upsampling of decompressed financial time-series data and mitigating compression artifacts.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of energy

If lossy compression is applied to financial time-series data, then bandwidth and storage efficiency are improved, but data accuracy and information completeness deteriorate

Engineering Contradiction:

Improvebandwidth usageVSAvoiddata accuracy

Core Design Contradiction:

Loss of energyVSLoss of information

Solution Approach 1:

The neural network is trained in advance on pairs of original and lossily compressed financial time-series data to learn the mapping from compressed to original data. This preliminary training enables the network to predict and reconstruct lost information during decompression without requiring additional real-time computational resources

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

A neural network serves as an intermediary component between the lossy compression decoder and the final data output. The network takes decompressed data as input and produces reconstructed data as output, effectively mediating the transition from compressed to high-fidelity data representation

Inventive Principle:

Principle #24Intermediary (Mediator)

2Productivity

If traditional decompression methods are used, then processing speed is maintained, but compression artifacts and data quality deteriorate

Engineering Contradiction:

Improveprocessing speedVSAvoiddata quality

Core Design Contradiction:

ProductivityVSManufacturing precision

Solution Approach 1:

The patent replaces traditional mechanical/mathematical decompression algorithms with a neural network-based system. The neural network learns complex non-linear mappings from compressed to original data, substituting conventional signal processing methods with machine learning-based reconstruction

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

3Quantity of substance

If high-frequency financial data is stored without compression, then data completeness is maintained, but storage requirements and computational burden increase

Engineering Contradiction:

Improvedata completenessVSAvoidstorage requirements

Core Design Contradiction:

Quantity of substanceVSVolume of stationary object

Solution Approach 1:

The system changes the representation parameters of financial data by applying lossy compression that reduces data precision in exchange for smaller storage requirements. The neural network then restores these parameters to their original quality, enabling efficient storage without permanent data degradation

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS12093972B1Upsampling of decompressed financial time—series data using a neural network

Publication Date: 2024.09.17 ATOMBEAM TECH INC

AI summary

A system and methods for upsampling of decompressed financial time-series data after lossy compression using a neural network that integrates AI-based techniques to enhance compression quality. It incorporates a novel deep-learning neural network that upsamples decompressed data to restore information lost during lossy compression, taking advantage of cross-correlations between time-series data sets.