Unified ML SDK for Mobile Model Compression

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current mobile application development is hindered by the fragmentation of machine learning services, leading to high processing and memory requirements due to the need to load and execute numerous SDKs, which complicates development and makes it challenging to run complex machine-learned models on resource-constrained devices.

Innovation Solution

A unified application development platform and SDK that provides a cross-platform API for machine learning services, enabling developers to generate, deploy, and manage machine-learned models, with features like model compression and conversion to optimize models for resource-constrained devices, using techniques such as quantization and pruning, and an end-to-end learning framework for creating compact, efficient models.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If multiple separate SDKs are used to provide machine learning services, then comprehensive machine learning functionality is achieved, but processing requirements and memory usage increase significantly

Engineering Contradiction:
Improvemachine learning functionalityVSAvoidprocessing requirements
Core Design Contradiction:
Adaptability or versatilityVSUse of energy by moving object

Solution Approach 1:

The patent combines multiple separate machine learning SDKs into a single unified SDK that provides comprehensive machine learning services. This consolidation merges previously separate functionality into one integrated package, reducing the need to load and execute multiple separate SDKs while maintaining full machine learning capabilities.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The unified SDK is designed to perform multiple machine learning functions within a single package, making it a universal solution that replaces several specialized SDKs. This multi-functional approach allows the single SDK to handle various machine learning tasks without requiring separate software packages for each function.

Inventive Principle:
Principle #6Universality (Multi-functionality)

2Adaptability or versatility

If multiple separate SDKs are used to provide machine learning services, then comprehensive machine learning functionality is achieved, but memory usage increases due to loading numerous libraries

Engineering Contradiction:
Improvemachine learning functionalityVSAvoidmemory usage
Core Design Contradiction:
Adaptability or versatilityVSQuantity of substance

Solution Approach 1:

The patent combines multiple separate machine learning SDKs into a single unified SDK that provides comprehensive machine learning services. This consolidation merges previously separate functionality into one integrated package, reducing the need to load and execute multiple separate SDKs while maintaining full machine learning capabilities.

Inventive Principle:
Principle #5Merging (Combining)

3Measurement precision

If complex machine-learned models are deployed on resource-constrained mobile devices, then model accuracy is maintained, but device resources are overwhelmed

Engineering Contradiction:
Improvemodel accuracyVSAvoidresource constraints
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent applies model compression techniques that modify parameters of machine learning models to reduce their size and computational requirements. By changing parameters such as precision (e.g., from float32 to float16 or int8) and applying pruning to remove unnecessary connections, the models can maintain acceptable accuracy while becoming suitable for deployment on resource-constrained mobile devices.

Inventive Principle:
Principle #35Parameter changes

4Adaptability or versatility

If multiple SDKs are integrated into a mobile application, then machine learning services are comprehensive, but development complexity increases

Engineering Contradiction:
Improvemachine learning servicesVSAvoiddevelopment complexity
Core Design Contradiction:
Adaptability or versatilityVSEase of manufacture

Solution Approach 1:

The patent combines multiple separate machine learning SDKs into a single unified SDK that provides comprehensive machine learning services. This consolidation merges previously separate functionality into one integrated package, reducing the need to load and execute multiple separate SDKs while maintaining full machine learning capabilities.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The unified SDK is designed to perform multiple machine learning functions within a single package, making it a universal solution that replaces several specialized SDKs. This multi-functional approach allows the single SDK to handle various machine learning tasks without requiring separate software packages for each function.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Data Source

PatentUS20240403031A1Application Development Platform and Software Development Kits that Provide Comprehensive Machine Learning Services
Publication Date: 2024.12.05 GOOGLE LLC
  • US20240403031A1 patent drawing
  • US20240403031A1 patent drawing
  • US20240403031A1 patent drawing

AI summary

The present disclosure provides an application development platform and associated software development kits (“SDKs”) that provide comprehensive services for generation, deployment, and management of machine-learned models used by computer applications such as, for example, mobile applications executed by a mobile computing device. In particular, the application development platform and SDKs can provide or otherwise leverage a unified, cross-platform application programming interface (“API”) that enables access to all of the different machine learning services needed for full machine learning functionality within the application. In such fashion, developers can have access to a single SDK for all machine learning services.