本文为加拿大多伦多大学（作者：Wei Cui）的硕士论文，共74页。

密集无线网络中干扰链路的优化调度是一个具有挑战性的课题。传统的方法是先估计所有干扰信道强度，然后根据模型优化调度，这两种方法都是计算密集型的。本论文探讨使用深度学习来进行调度的方法，是完全基于地理位置的输入。我们通过使用一种新的神经网络结构来创新解决方案，该结构通过多个反馈阶段计算用于干扰估计的地理空间卷积，通过无监督训练来学习最佳解。所得到的神经网络在和速率最大化方面具有近似最优的性能，并且能够推广到更大的问题规模。此外，本文还致力于优化整个网络的比例公平性，提出了一种新的调度方法，利用和速率优化调度启发式算法对经过明智选择的链路子集进行优化，显示出了很强的竞争性能。

The optimal scheduling of interfering links in a dense wireless network with full frequency reuse is a challenging task. The traditional method involves first estimating all the interfering channel strengths thenoptimizing the scheduling based on the model, both of which are computationally intensive. This thesisexplores approaches using deep learning for scheduling, based solely on the geographic locations inputs.We innovated a solution by using a novel neural network architecture that computes the geographicspatial convolutions for interference estimation, over multiple feedback stages to learn the optimum solution through unsupervised training. The resulting neural network gives near-optimal performance forsum-rate maximization and is capable of generalizing to larger problem sizes. Furthermore, this thesisalso focuses on optimizing Proportional Fairness across the network, through proposing a novel scheduling approach that utilizes the sum-rate optimal scheduling heuristics over judiciously chosen subsets oflinks, showing highly competitive performances.

1.  引言

2.  问题描述

3.  已有的方法

4.  采用卷积神经网络结构的动机

5.  训练方法

6.  和速率优化的创新卷积神经网络模型

7.  对数效用优化：加权和速率最大化的挑战

8.  对数效用优化：创新的解决方案

9.  仿真结果：和速率优化

10.          仿真结果：对数效用优化

11.          结论

更多精彩文章请关注公众号：