Optical Data Preprocessing With Overlapped Memory for AI Bottlenecks

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current computing architectures face energy inefficiencies and memory bottlenecks during large-scale data processing, leading to delays and data loss in AI operations, particularly in image processing tasks.

Innovation Solution

An optical data preprocessor using optical memory devices for overlapping and storing optical signals, reducing data transmission and processing time by converting and storing input data as overlapped image data, utilizing photoelectric conversion layers with suppressed spike-time dependent plasticity and maintained spike-number dependent plasticity characteristics.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Use of energy by moving object

If conventional von Neumann-based computing architectures are used for AI operations, then processing capability is maintained, but energy consumption increases significantly and scalability is limited

Engineering Contradiction:
Improveenergy consumptionVSAvoidprocessing capability
Core Design Contradiction:
Use of energy by moving objectVSProductivity

Solution Approach 1:

The patent replaces conventional von Neumann-based computing architectures with optical computing systems that use light instead of electrical signals for data processing. This substitution fundamentally changes the physical basis of computation, enabling parallel processing of optical signals through photodetectors and optical modulators, thereby reducing energy consumption while maintaining or enhancing processing capability for AI operations

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent merges memory and processing functions into a unified optical computing system where optical signals are directly processed without conversion to electrical domain. The optical memory device stores optical signals, and the optical modulator processes them in-place, eliminating the separation between memory and processor that characterizes von Neumann architectures, thus reducing energy consumption and improving scalability

Inventive Principle:
Principle #5Merging (Combining)

2Productivity

If preprocessing reduces the number of pixels by lowering image resolution, then processing speed improves, but data loss occurs and image quality degrades

Engineering Contradiction:
Improveprocessing speedVSAvoiddata loss
Core Design Contradiction:
ProductivityVSLoss of information

Solution Approach 1:

The patent performs preliminary processing of optical signals directly in the optical domain before conversion to electrical signals. By using optical modulators to adjust signal characteristics and optical memory to store intermediate results, the system prepares data in a way that maintains information integrity while enabling faster subsequent processing, thus achieving speed improvement without significant data loss

Inventive Principle:
Principle #10Preliminary action

3Quantity of substance

If analog signals are converted to digital format and stored in non-volatile memory for processing, then data storage capability is improved, but processing complexity increases and power loss occurs due to large-scale data transmission

Engineering Contradiction:
Improvedata storage capacityVSAvoidprocessing complexity
Core Design Contradiction:
Quantity of substanceVSDevice complexity

Solution Approach 1:

The patent extracts only the essential features and characteristics of optical signals for storage and processing, rather than converting and storing complete high-resolution digital representations. By using optical memory to store compressed or feature-extracted data and processing only this reduced dataset, the system achieves adequate storage capacity while significantly reducing processing complexity and transmission power requirements

Inventive Principle:
Principle #2Taking out (Extraction)

4Loss of information

If all analog data are transmitted for processing, then data completeness is maintained, but power consumption increases due to large-scale data transfer

Engineering Contradiction:
Improvedata completenessVSAvoidpower consumption
Core Design Contradiction:
Loss of informationVSUse of energy by moving object

Solution Approach 1:

The patent extracts and processes only the most relevant features from optical signals using optical modulators and selective filtering in the optical domain. By identifying and transmitting only essential data elements rather than complete analog datasets, the system maintains sufficient information completeness for accurate processing while dramatically reducing the volume of data transmission and associated power consumption

Inventive Principle:
Principle #2Taking out (Extraction)

5Quantity of substance

If memory bottlenecks are present between memory and processing unit, then data storage is sufficient, but processing delays occur

Engineering Contradiction:
Improvedata storageVSAvoidprocessing delay
Core Design Contradiction:
Quantity of substanceVSLoss of time

Solution Approach 1:

The patent merges memory and processing functions into an integrated optical computing system where optical memory devices are directly coupled with optical modulators and photodetectors. This integration eliminates the physical and functional separation between memory and processor, allowing data to be processed in-place without transfer delays, thus resolving memory bottlenecks while maintaining adequate storage capacity

Inventive Principle:
Principle #5Merging (Combining)

Applied Scientific Principles

This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.

Function Achieved in This Case

This approach significantly reduces data volume, minimizes memory bottlenecks, and enhances processing speed and accuracy in AI training and inference, particularly in computer vision applications like autonomous driving and IoT, while lowering power consumption.

Implementation Method 1

Each of the optical memory devices may include: a substrate; a photoelectric conversion layer stacked on the substrate; and a source electrode and a drain electrode spaced apart from each other and formed on the photoelectric conversion layer.

Methodology Applied
Scientific EffectPhotoelectric conversion: Photoelectric Effect

Implementation Method 2

Each of the optical memory devices constituting the optical memory device array may have a photoelectric conversion and memory function that stores the optically input image data in an overlapped manner based on a conductivity change according to the number of times the optical image data are input.

Methodology Applied
Scientific EffectConductivity change accumulation:

Implementation Method 3

The photoelectric conversion layer may exhibit suppressed spike-time dependent plasticity (STDP) characteristics and exhibit spike-number dependent plasticity (SNDP) characteristics, so that the conductivity changes are accumulated with the same weight for each of the optically input image data that are input in a time-series manner.

Methodology Applied
Scientific EffectSpike-number dependent plasticity (SNDP):

Data Source

PatentUS20250363587A1Optical data preprocessor based on optical neuromorphic device
Publication Date: 2025.11.27 UNIVERSITY INDUSTRY COOPERATION GROUP OF KYUNG HEE UNIVERSITY
  • US20250363587A1 patent drawing
  • US20250363587A1 patent drawing
  • US20250363587A1 patent drawing

AI summary

The present invention provides an optical memory device-based optical data preprocessor and a method thereof, which improves data processing speed and reduces data bottlenecks by storing input optical signals in an overlapped manner and then outputting them as electrical signals for data processing.