Deep Learning
文章平均质量分 89
Ravi-Jay
自知在我
展开
-
cs231n-notes-Lecture-2/3
CS231n主要是关于image classification的一些算法以及迁移到其他问题的几个应用。这几节的内容比较基础,正好有时间就过一遍。Lecture 2 image classification pipelinecompare two imagesL1 (Mahanttan)Distance: $ L = \sum{|X_{ij} - Y_{ij}|}$depends on ...原创 2018-09-18 14:52:40 · 123 阅读 · 0 评论 -
cs231n-notes-Lecture-4/5/6: 反向传播/激活函数/数据预处理/权重初始化/batch norm
Lecture-4 Backpropagation and Neural NetworksComputational GraphsNode gradient = [local gradient] x [upstream gradient]add gate: gradient distributormax gate: gradient router (choose only a way)...原创 2018-09-18 15:00:51 · 280 阅读 · 0 评论 -
cs231n-notes-Lecture-7:各种优化方法介绍与比较
Lecture-7 Training Neural NetworksOptimizationSGDConsVery slow progress along shallow dimension, jitter along steep direction. 2. local minima or saddle point. Saddle points are much more comm...原创 2018-09-18 15:07:53 · 371 阅读 · 0 评论 -
cs231n-notes-Lecture-9:AlexNet/VGGNet/GoogleNet/ResNet
Lecture 9 CNN ArchitecturesAlexNetVGGNetSmall filtersWhy?Stack of three 3x3 conv (stride 1) layers has same effective receptive field as one 7x7 conv layerBut deeper, more non-linearitiesAn...原创 2018-09-18 15:11:22 · 235 阅读 · 0 评论 -
cs231n-notes-Lecture-10:Recurrent Neural Networks
Recurrent Neural Networks: Process Sequencesone to one: Vanilla Neural Networks. rawone to many: eg. Image Captioning image -> sequence of wordsmany to one: e.g. Sentiment Classification seque...原创 2018-09-18 15:15:14 · 197 阅读 · 0 评论