深度学习相关概念知识链接（自用

腾讯云作者：本立2道生
网络权重初始化方法总结（上）：梯度消失、梯度爆炸与不良的初始化
https://cloud.tencent.com/developer/article/1535198
网络权重初始化方法总结（下）：Lecun、Xavier与He Kaiming
https://cloud.tencent.com/developer/article/1542736

为什么要做特征归一化/标准化？
https://cloud.tencent.com/developer/article/1531996

Batch Normalization详解
https://cloud.tencent.com/developer/article/1551518
直观理解为什么分类问题用交叉熵损失而不用均方误差损失?
https://cloud.tencent.com/developer/article/1554408
特征，特征不变性，尺度空间与图像金字塔
https://cloud.tencent.com/developer/article/1374374
数字图像处理，计算机视觉，计算机图形学，计算摄影
https://cloud.tencent.com/developer/article/1583857
直观理解神经网络最后一层全连接+Softmax
https://cloud.tencent.com/developer/article/1376756
ImageNet主要网络benchmark对比
https://cloud.tencent.com/developer/article/1495763
Classifying with k-Nearest Neighbors（k近邻）
https://cloud.tencent.com/developer/article/1505585
从AlexNet(2012)开始
https://cloud.tencent.com/developer/article/1505678
Network in Network(2013)，1x1卷积与Global Average Pooling
https://cloud.tencent.com/developer/article/1520061
VGG(2014)，3x3卷积的胜利
https://cloud.tencent.com/developer/article/1521436
ResNet详解与分析
https://cloud.tencent.com/developer/article/1591484
通俗易懂DenseNet
https://cloud.tencent.com/developer/article/1594171
Inception系列理解
https://cloud.tencent.com/developer/article/1610944