Machine Learning Models for a Novel Optical Memory Approach

Area of Science:

• Optoelectronics and data storage solutions.
• Nanoparticle-based optical memory.
• Machine learning applications in photonics.

Background:

• Growing demand for high-speed, high-bandwidth data handling in data centers.
• Need for efficient nonvolatile optical memory for all-optical data processing.
• Previous introduction of optical memory based on gold nanoparticle scattering fields.

Purpose of the Study:

• To enhance the performance of a nonvolatile optical memory element.
• To apply advanced machine learning techniques for improved classification of optical data.
• To analyze scattered images from the optical memory device using sophisticated algorithms.

Main Methods:

• Utilizing Random Forest and t-SNE (t-distributed Stochastic Neighbor Embedding) algorithms.
• Classification and analysis of scattered images generated by the optical memory device.
• Development of a machine learning model for optical data classification.

Main Results:

• Successful classification and analysis of scattered images from the optical memory device.
• Achieved an accuracy of 0.81 for the classification model.
• Attained an average F1-score of 0.81 across nine distinct classes.

Conclusions:

• Machine learning techniques significantly enhance nonvolatile optical memory performance.
• The proposed classification model demonstrates high accuracy in distinguishing optical data classes.
• This research advances the development of efficient all-optical data processing systems.