# Distributed Deep Learning at the Edge: A Novel Proactive and Cooperative Caching Framework for Mobile Edge Networks

## TL;DR

This paper introduces two deep learning-based proactive caching strategies for mobile edge networks, improving content demand prediction accuracy and reducing service delay while addressing privacy concerns.

## Contribution

It presents a novel distributed deep learning framework for cooperative caching that preserves user privacy and enhances prediction accuracy in edge networks.

## Key findings

- Prediction accuracy improved by up to 33.7% RMSE reduction.
- Service delay decreased by 36.1%.
- Distributed approach outperforms centralized methods.

## Abstract

This letter proposes two novel proactive cooperative caching approaches using deep learning (DL) to predict users' content demand in a mobile edge caching network. In the first approach, a (central) content server takes responsibilities to collect information from all mobile edge nodes (MENs) in the network and then performs our proposed deep learning (DL) algorithm to predict the content demand for the whole network. However, such a centralized approach may disclose the private information because MENs have to share their local users' data with the content server. Thus, in the second approach, we propose a novel distributed deep learning (DDL) based framework. The DDL allows MENs in the network to collaborate and exchange information to reduce the error of content demand prediction without revealing the private information of mobile users. Through simulation results, we show that our proposed approaches can enhance the accuracy by reducing the root mean squared error (RMSE) up to 33.7% and reduce the service delay by 36.1% compared with other machine learning algorithms.