Projection Neural Network for Local Processing

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional neural networks require significant computational resources and memory, making them unsuitable for deployment in resource-constrained systems, and they often transmit data remotely for processing, which compromises data privacy.

Innovation Solution

The implementation of a projection neural network with a sequence of projection layers that reduce input dimensionality using sparse matrices and bias vectors, allowing for fewer parameters and arithmetic operations, enabling efficient local processing in resource-constrained environments.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If conventional neural networks are deployed, then processing capability is improved, but computational resource consumption increases

Engineering Contradiction:

Improveprocessing capabilityVSAvoidcomputational resource consumption

Core Design Contradiction:

ProductivityVSUse of energy by moving object

Solution Approach 1:

The patent segments the neural network into projection layers that operate independently to reduce dimensionality. Each projection layer processes data in smaller dimensional spaces, dividing the overall computational task into manageable segments that consume fewer resources while maintaining processing capability.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent transforms data from high-dimensional space to low-dimensional space through projection layers. By changing the dimensionality of the data representation, the network reduces computational complexity and resource consumption while preserving essential information needed for processing.

Inventive Principle:

Principle #17Another dimension (Dimensionality change)

2Productivity

If conventional neural networks are deployed, then processing capability is improved, but device complexity increases

Engineering Contradiction:

Improveprocessing capabilityVSAvoidnetwork complexity

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The network is segmented into simple projection layers with straightforward mathematical operations rather than complex interconnected neurons. This segmentation simplifies the overall network structure while maintaining processing capability through sequential dimensionality reduction.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent replaces the traditional mechanical neural network structure with projection-based mathematical transformations. This substitution simplifies the system by using linear algebra operations instead of complex neuron interactions, reducing device complexity while preserving functionality.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

3Productivity

If data is transmitted remotely for processing, then processing capability is improved, but data privacy deteriorates

Engineering Contradiction:

Improveprocessing capabilityVSAvoiddata privacy risk

Core Design Contradiction:

ProductivityVSObject-affected harmful factors

Solution Approach 1:

The patent enables the device to perform processing locally using the projection neural network, making the system self-sufficient. Data remains on the device throughout processing, eliminating the need for remote transmission and thereby protecting data privacy while maintaining processing capability.

Inventive Principle:

Principle #25Self-service

Data Source

PatentUS11544573B2Projection neural networks

Publication Date: 2023.01.03 GOOGLE LLC

AI summary

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for a projection neural network. In one aspect, a projection neural network is configured to receive a projection network input and to generate a projection network output from the projection network input. The projection neural network includes a sequence of one or more projection layers. Each projection layer has multiple projection layer parameters, and is configured to receive a layer input, apply multiple projection layer functions to the layer input, and generate a layer output by applying the projection layer parameters for the projection layer to the projection function outputs.