-
Notifications
You must be signed in to change notification settings - Fork 0
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
- Loading branch information
Showing
1 changed file
with
23 additions
and
1 deletion.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -1 +1,23 @@ | ||
# DL_intorduction | ||
# DL_intorduction | ||
|
||
## 第一讲 | ||
不讲(默认掌握):loss, SGD, optimizer, lr, MLP, softmax... | ||
今日内容概要: | ||
1. 基于pytorch实现一个cifar10分类任务的baseline model | ||
- 特征工程初探(基于transforms的standarlize,Augmentation),图像变换 | ||
- 模型搭建(前向反向传播,Trainer Demo,卷积核,池化,norm) | ||
- Optimizer(SGD,Adam,Weight decay,余弦退火原理) | ||
2. 对Bottleneck优化与进一步探究,加深理解 | ||
- 随着网络层数加深,模型效果是否一定变好?(Resnet,跳跃连接) | ||
- 玄学调参设置:不同任务lr该是多少?batch选择对模型准确率影响?(我的经验) | ||
- 如何对数据做增强,使得小数据集上一样获得较高的模型性能?(甚至Datafree) | ||
- 如何让模型快速收敛?模型前后层如何保持学到的特征不遗忘?(Norm) | ||
- 玄学调参2:模型参数初始化设置?(Kaiming initialize & Xavier) | ||
- 同等网络层数下如何提取更多特征?(通道数翻倍?多头注意力?) | ||
- 玄学解释3:为什么神经网络可以work?(feature map的可解释性) | ||
3. 有空再讲: | ||
- 如何在该数据集上达到最高准确率?尝试逼近SOTA | ||
- 有无办法,给定数据集后不写任何神经网络代码,自动完成该问题下模型架构搭建与超参数设置(太伟大了NAS) | ||
- 模型太笨重,推理速度慢?(太伟大了模型压缩,剪枝量化,甚至可以部署在单片机上) | ||
- 如何让另一个模型学到一个已经训练好的模型的知识?(无需重新训练,太伟大了模型蒸馏) | ||
- 更多更多的trick分享.....以及一点深度学习的哲学思考 |