Convolutional neural network-based data page classification for holographic memory.

Area of Science:

• Computer Science
• Optics and Photonics
• Data Storage Technologies

Background:

• Holographic memory offers high storage density but faces challenges with data page classification.
• Noise and lateral shifts in reconstructed data pages can degrade classification accuracy.
• Conventional methods like multilayer perceptrons (MLPs) may struggle with complex holographic data.

Purpose of the Study:

• To propose and evaluate a deep-learning-based method for classifying data pages in holographic memory.
• To compare the performance of a deep neural network (DNN) against a conventional MLP under noisy and shifted conditions.
• To quantify the accuracy improvements offered by DNNs for holographic data retrieval.

Main Methods:

• Numerical investigation of classification performance using a multilayer perceptron (MLP) and a deep neural network (DNN).
• Simulated holographic data pages contaminated with noise and subjected to random lateral shifts.
• Evaluation of classification accuracy for both MLP and DNN under these challenging conditions.

Main Results:

• The multilayer perceptron achieved a classification accuracy of 93.02% for laterally shifted data pages.
• The deep neural network demonstrated a significantly higher classification accuracy of 99.98%.
• The DNN's accuracy was approximately two orders of magnitude better than the MLP's, highlighting its robustness.

Conclusions:

• Deep neural networks provide a superior solution for accurate data page classification in holographic memory systems.
• The proposed DNN approach effectively mitigates the impact of noise and lateral shifts, crucial for reliable holographic data storage.
• This advancement paves the way for more robust and high-performance holographic memory technologies.