Edge AI Training Compression Using Cluster-Based Surrogate Data

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Edge computing environments face performance constraints and security threats due to data transfer inefficiencies and adversarial data poisoning, which affect the stability and reliability of AI models in distributed systems.

Innovation Solution

Implementing a lossy compression algorithm in edge devices to generate compressed encodings of numeric values from IoT units, which are then used to create surrogates for training AI models, and employing clustering algorithms to identify and filter out adversarial data, thereby enhancing data communication efficiency and model robustness.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If lossy compression algorithm is applied to reduce data communication, then data transfer efficiency is improved, but measurement precision of numeric values deteriorates

Engineering Contradiction:

Improvedata transfer efficiencyVSAvoidnumeric value precision

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The patent changes the parameter representation from high-precision floating-point numbers to low-precision cluster indices. Numeric values are transformed by mapping them to cluster centers, effectively changing the precision parameter from high to low while enabling efficient compression and transmission.

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

Instead of transmitting the actual numeric values, the system transmits compressed encodings (cluster indices) that represent groups of similar values. The cloud then generates surrogate values by sampling from probability distributions associated with each cluster, creating copies that preserve statistical properties without requiring exact value transmission.

Inventive Principle:

Principle #26Copying

2Loss of substance

If clustering algorithm is used to compress data, then data communication volume is reduced, but information loss increases

Engineering Contradiction:

Improvedata communication volumeVSAvoiddata information

Core Design Contradiction:

Loss of substanceVSLoss of information

Solution Approach 1:

The patent introduces probability distributions as intermediaries between the compressed cluster indices and the original data values. These distributions act as mediators that preserve statistical information (mean, variance) while enabling efficient compression. The surrogate values are sampled from these distributions, recovering the statistical properties of the original data without transmitting the actual values.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The system changes the information representation from individual data values to statistical parameters (cluster indices, probability distributions). This parameter transformation reduces communication volume while preserving the essential statistical characteristics needed for AI training, accepting that exact individual values are lost but population-level properties are maintained.

Inventive Principle:

Principle #35Parameter changes

3Reliability

If surrogate values are generated for AI training, then model robustness is improved, but training data accuracy deteriorates

Engineering Contradiction:

Improvemodel robustnessVSAvoidtraining data accuracy

Core Design Contradiction:

ReliabilityVSMeasurement precision

Solution Approach 1:

The patent converts the potential harm of data compression (information loss) into a benefit by showing that surrogate values generated from probability distributions improve model robustness. The random sampling process introduces beneficial noise that prevents overfitting, while the compression and surrogate generation process filters out adversarial perturbations, converting data imperfections into training advantages.

Inventive Principle:

Principle #22Blessing in disguise (Convert harm into benefit)

Solution Approach 2:

The system implements a feedback mechanism where cluster statistics (mean, variance) are computed from historical data and used to generate probability distributions. These distributions then generate surrogate values for training, creating a closed-loop system where past data informs future training data generation, continuously improving model robustness while maintaining compression efficiency.

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS11728825B2Cluster-based data compression for AI training on the cloud for an edge network

Publication Date: 2023.08.15 DELL PROD LP

AI summary

A disclosed information handling system includes an edge device communicatively coupled to a cloud computing resource. The edge device is configured to respond to receiving, from an internet of things (IoT) unit, a numeric value for a parameter of interest by determining a compressed encoding for the numeric value in accordance with a non-lossless compression algorithm. The edge device transmits the compressed encoding of the numeric value to the cloud computing resource. The cloud computing resource includes a decoder communicatively coupled to the encoder and configured to respond to receiving the compressed encoding by generating a surrogate for the numeric value. The surrogate may be generated in accordance with a probability distribution applicable to the parameter of interest. The compression algorithm may be a clustering algorithm such as a k-means clustering algorithm.